Delete benchmark directory

benchmark/csv/meta_bonn.csv

@@ -1,6 +0,0 @@
-filepath_left,filepath_disparity
-bonn/rgbd_bonn_synchronous_rgb_left.mp4,bonn/rgbd_bonn_synchronous_disparity.npz
-bonn/rgbd_bonn_person_tracking_rgb_left.mp4,bonn/rgbd_bonn_person_tracking_disparity.npz
-bonn/rgbd_bonn_crowd2_rgb_left.mp4,bonn/rgbd_bonn_crowd2_disparity.npz
-bonn/rgbd_bonn_crowd3_rgb_left.mp4,bonn/rgbd_bonn_crowd3_disparity.npz
-bonn/rgbd_bonn_balloon2_rgb_left.mp4,bonn/rgbd_bonn_balloon2_disparity.npz

benchmark/csv/meta_kitti_val.csv

@@ -1,14 +0,0 @@
-filepath_left,filepath_disparity
-KITTI/2011_09_28_drive_0037_sync_rgb_left.mp4,KITTI/2011_09_28_drive_0037_sync_disparity.npz
-KITTI/2011_09_26_drive_0005_sync_rgb_left.mp4,KITTI/2011_09_26_drive_0005_sync_disparity.npz
-KITTI/2011_09_30_drive_0016_sync_rgb_left.mp4,KITTI/2011_09_30_drive_0016_sync_disparity.npz
-KITTI/2011_09_26_drive_0079_sync_rgb_left.mp4,KITTI/2011_09_26_drive_0079_sync_disparity.npz
-KITTI/2011_09_26_drive_0020_sync_rgb_left.mp4,KITTI/2011_09_26_drive_0020_sync_disparity.npz
-KITTI/2011_09_26_drive_0095_sync_rgb_left.mp4,KITTI/2011_09_26_drive_0095_sync_disparity.npz
-KITTI/2011_10_03_drive_0047_sync_rgb_left.mp4,KITTI/2011_10_03_drive_0047_sync_disparity.npz
-KITTI/2011_09_26_drive_0113_sync_rgb_left.mp4,KITTI/2011_09_26_drive_0113_sync_disparity.npz
-KITTI/2011_09_26_drive_0036_sync_rgb_left.mp4,KITTI/2011_09_26_drive_0036_sync_disparity.npz
-KITTI/2011_09_26_drive_0013_sync_rgb_left.mp4,KITTI/2011_09_26_drive_0013_sync_disparity.npz
-KITTI/2011_09_26_drive_0002_sync_rgb_left.mp4,KITTI/2011_09_26_drive_0002_sync_disparity.npz
-KITTI/2011_09_29_drive_0026_sync_rgb_left.mp4,KITTI/2011_09_29_drive_0026_sync_disparity.npz
-KITTI/2011_09_26_drive_0023_sync_rgb_left.mp4,KITTI/2011_09_26_drive_0023_sync_disparity.npz

benchmark/csv/meta_nyu_test.csv

@@ -1,655 +0,0 @@
-filepath_left,filepath_disparity
-NYUv2/test/kitchen_0004/rgb_0001_rgb_left.mp4,NYUv2/test/kitchen_0004/rgb_0001_disparity.npz
-NYUv2/test/kitchen_0004/rgb_0002_rgb_left.mp4,NYUv2/test/kitchen_0004/rgb_0002_disparity.npz
-NYUv2/test/office_0005/rgb_0009_rgb_left.mp4,NYUv2/test/office_0005/rgb_0009_disparity.npz
-NYUv2/test/office_0007/rgb_0014_rgb_left.mp4,NYUv2/test/office_0007/rgb_0014_disparity.npz
-NYUv2/test/office_0008/rgb_0015_rgb_left.mp4,NYUv2/test/office_0008/rgb_0015_disparity.npz
-NYUv2/test/office_0008/rgb_0016_rgb_left.mp4,NYUv2/test/office_0008/rgb_0016_disparity.npz
-NYUv2/test/office_0008/rgb_0017_rgb_left.mp4,NYUv2/test/office_0008/rgb_0017_disparity.npz
-NYUv2/test/office_0008/rgb_0018_rgb_left.mp4,NYUv2/test/office_0008/rgb_0018_disparity.npz
-NYUv2/test/office_0010/rgb_0021_rgb_left.mp4,NYUv2/test/office_0010/rgb_0021_disparity.npz
-NYUv2/test/office_0013/rgb_0028_rgb_left.mp4,NYUv2/test/office_0013/rgb_0028_disparity.npz
-NYUv2/test/office_0013/rgb_0029_rgb_left.mp4,NYUv2/test/office_0013/rgb_0029_disparity.npz
-NYUv2/test/office_0013/rgb_0030_rgb_left.mp4,NYUv2/test/office_0013/rgb_0030_disparity.npz
-NYUv2/test/office_0013/rgb_0031_rgb_left.mp4,NYUv2/test/office_0013/rgb_0031_disparity.npz
-NYUv2/test/office_0013/rgb_0032_rgb_left.mp4,NYUv2/test/office_0013/rgb_0032_disparity.npz
-NYUv2/test/office_0013/rgb_0033_rgb_left.mp4,NYUv2/test/office_0013/rgb_0033_disparity.npz
-NYUv2/test/office_0013/rgb_0034_rgb_left.mp4,NYUv2/test/office_0013/rgb_0034_disparity.npz
-NYUv2/test/office_0014/rgb_0035_rgb_left.mp4,NYUv2/test/office_0014/rgb_0035_disparity.npz
-NYUv2/test/office_0014/rgb_0036_rgb_left.mp4,NYUv2/test/office_0014/rgb_0036_disparity.npz
-NYUv2/test/office_0014/rgb_0037_rgb_left.mp4,NYUv2/test/office_0014/rgb_0037_disparity.npz
-NYUv2/test/office_0014/rgb_0038_rgb_left.mp4,NYUv2/test/office_0014/rgb_0038_disparity.npz
-NYUv2/test/office_0015/rgb_0039_rgb_left.mp4,NYUv2/test/office_0015/rgb_0039_disparity.npz
-NYUv2/test/office_0015/rgb_0040_rgb_left.mp4,NYUv2/test/office_0015/rgb_0040_disparity.npz
-NYUv2/test/office_0015/rgb_0041_rgb_left.mp4,NYUv2/test/office_0015/rgb_0041_disparity.npz
-NYUv2/test/office_0015/rgb_0042_rgb_left.mp4,NYUv2/test/office_0015/rgb_0042_disparity.npz
-NYUv2/test/office_0015/rgb_0043_rgb_left.mp4,NYUv2/test/office_0015/rgb_0043_disparity.npz
-NYUv2/test/bathroom_0003/rgb_0046_rgb_left.mp4,NYUv2/test/bathroom_0003/rgb_0046_disparity.npz
-NYUv2/test/bathroom_0004/rgb_0047_rgb_left.mp4,NYUv2/test/bathroom_0004/rgb_0047_disparity.npz
-NYUv2/test/bedroom_0011/rgb_0056_rgb_left.mp4,NYUv2/test/bedroom_0011/rgb_0056_disparity.npz
-NYUv2/test/bedroom_0011/rgb_0057_rgb_left.mp4,NYUv2/test/bedroom_0011/rgb_0057_disparity.npz
-NYUv2/test/bedroom_0013/rgb_0059_rgb_left.mp4,NYUv2/test/bedroom_0013/rgb_0059_disparity.npz
-NYUv2/test/bedroom_0013/rgb_0060_rgb_left.mp4,NYUv2/test/bedroom_0013/rgb_0060_disparity.npz
-NYUv2/test/bedroom_0013/rgb_0061_rgb_left.mp4,NYUv2/test/bedroom_0013/rgb_0061_disparity.npz
-NYUv2/test/bedroom_0013/rgb_0062_rgb_left.mp4,NYUv2/test/bedroom_0013/rgb_0062_disparity.npz
-NYUv2/test/bedroom_0013/rgb_0063_rgb_left.mp4,NYUv2/test/bedroom_0013/rgb_0063_disparity.npz
-NYUv2/test/bedroom_0018/rgb_0076_rgb_left.mp4,NYUv2/test/bedroom_0018/rgb_0076_disparity.npz
-NYUv2/test/bedroom_0018/rgb_0077_rgb_left.mp4,NYUv2/test/bedroom_0018/rgb_0077_disparity.npz
-NYUv2/test/bedroom_0018/rgb_0078_rgb_left.mp4,NYUv2/test/bedroom_0018/rgb_0078_disparity.npz
-NYUv2/test/bedroom_0018/rgb_0079_rgb_left.mp4,NYUv2/test/bedroom_0018/rgb_0079_disparity.npz
-NYUv2/test/bookstore_0001/rgb_0084_rgb_left.mp4,NYUv2/test/bookstore_0001/rgb_0084_disparity.npz
-NYUv2/test/bookstore_0001/rgb_0085_rgb_left.mp4,NYUv2/test/bookstore_0001/rgb_0085_disparity.npz
-NYUv2/test/bookstore_0001/rgb_0086_rgb_left.mp4,NYUv2/test/bookstore_0001/rgb_0086_disparity.npz
-NYUv2/test/bookstore_0001/rgb_0087_rgb_left.mp4,NYUv2/test/bookstore_0001/rgb_0087_disparity.npz
-NYUv2/test/bookstore_0001/rgb_0088_rgb_left.mp4,NYUv2/test/bookstore_0001/rgb_0088_disparity.npz
-NYUv2/test/bookstore_0001/rgb_0089_rgb_left.mp4,NYUv2/test/bookstore_0001/rgb_0089_disparity.npz
-NYUv2/test/bookstore_0001/rgb_0090_rgb_left.mp4,NYUv2/test/bookstore_0001/rgb_0090_disparity.npz
-NYUv2/test/bookstore_0001/rgb_0091_rgb_left.mp4,NYUv2/test/bookstore_0001/rgb_0091_disparity.npz
-NYUv2/test/bookstore_0001/rgb_0117_rgb_left.mp4,NYUv2/test/bookstore_0001/rgb_0117_disparity.npz
-NYUv2/test/bookstore_0001/rgb_0118_rgb_left.mp4,NYUv2/test/bookstore_0001/rgb_0118_disparity.npz
-NYUv2/test/bookstore_0001/rgb_0119_rgb_left.mp4,NYUv2/test/bookstore_0001/rgb_0119_disparity.npz
-NYUv2/test/kitchen_0005/rgb_0125_rgb_left.mp4,NYUv2/test/kitchen_0005/rgb_0125_disparity.npz
-NYUv2/test/kitchen_0005/rgb_0126_rgb_left.mp4,NYUv2/test/kitchen_0005/rgb_0126_disparity.npz
-NYUv2/test/kitchen_0005/rgb_0127_rgb_left.mp4,NYUv2/test/kitchen_0005/rgb_0127_disparity.npz
-NYUv2/test/kitchen_0005/rgb_0128_rgb_left.mp4,NYUv2/test/kitchen_0005/rgb_0128_disparity.npz
-NYUv2/test/kitchen_0005/rgb_0129_rgb_left.mp4,NYUv2/test/kitchen_0005/rgb_0129_disparity.npz
-NYUv2/test/kitchen_0007/rgb_0131_rgb_left.mp4,NYUv2/test/kitchen_0007/rgb_0131_disparity.npz
-NYUv2/test/kitchen_0007/rgb_0132_rgb_left.mp4,NYUv2/test/kitchen_0007/rgb_0132_disparity.npz
-NYUv2/test/kitchen_0007/rgb_0133_rgb_left.mp4,NYUv2/test/kitchen_0007/rgb_0133_disparity.npz
-NYUv2/test/kitchen_0007/rgb_0134_rgb_left.mp4,NYUv2/test/kitchen_0007/rgb_0134_disparity.npz
-NYUv2/test/kitchen_0009/rgb_0137_rgb_left.mp4,NYUv2/test/kitchen_0009/rgb_0137_disparity.npz
-NYUv2/test/living_room_0008/rgb_0153_rgb_left.mp4,NYUv2/test/living_room_0008/rgb_0153_disparity.npz
-NYUv2/test/living_room_0008/rgb_0154_rgb_left.mp4,NYUv2/test/living_room_0008/rgb_0154_disparity.npz
-NYUv2/test/living_room_0009/rgb_0155_rgb_left.mp4,NYUv2/test/living_room_0009/rgb_0155_disparity.npz
-NYUv2/test/living_room_0013/rgb_0167_rgb_left.mp4,NYUv2/test/living_room_0013/rgb_0167_disparity.npz
-NYUv2/test/living_room_0013/rgb_0168_rgb_left.mp4,NYUv2/test/living_room_0013/rgb_0168_disparity.npz
-NYUv2/test/living_room_0014/rgb_0169_rgb_left.mp4,NYUv2/test/living_room_0014/rgb_0169_disparity.npz
-NYUv2/test/bedroom_0003/rgb_0171_rgb_left.mp4,NYUv2/test/bedroom_0003/rgb_0171_disparity.npz
-NYUv2/test/bedroom_0003/rgb_0172_rgb_left.mp4,NYUv2/test/bedroom_0003/rgb_0172_disparity.npz
-NYUv2/test/bedroom_0003/rgb_0173_rgb_left.mp4,NYUv2/test/bedroom_0003/rgb_0173_disparity.npz
-NYUv2/test/bedroom_0003/rgb_0174_rgb_left.mp4,NYUv2/test/bedroom_0003/rgb_0174_disparity.npz
-NYUv2/test/bedroom_0003/rgb_0175_rgb_left.mp4,NYUv2/test/bedroom_0003/rgb_0175_disparity.npz
-NYUv2/test/bedroom_0003/rgb_0176_rgb_left.mp4,NYUv2/test/bedroom_0003/rgb_0176_disparity.npz
-NYUv2/test/bedroom_0005/rgb_0180_rgb_left.mp4,NYUv2/test/bedroom_0005/rgb_0180_disparity.npz
-NYUv2/test/bedroom_0005/rgb_0181_rgb_left.mp4,NYUv2/test/bedroom_0005/rgb_0181_disparity.npz
-NYUv2/test/bedroom_0005/rgb_0182_rgb_left.mp4,NYUv2/test/bedroom_0005/rgb_0182_disparity.npz
-NYUv2/test/bedroom_0006/rgb_0183_rgb_left.mp4,NYUv2/test/bedroom_0006/rgb_0183_disparity.npz
-NYUv2/test/bedroom_0006/rgb_0184_rgb_left.mp4,NYUv2/test/bedroom_0006/rgb_0184_disparity.npz
-NYUv2/test/bedroom_0006/rgb_0185_rgb_left.mp4,NYUv2/test/bedroom_0006/rgb_0185_disparity.npz
-NYUv2/test/bedroom_0006/rgb_0186_rgb_left.mp4,NYUv2/test/bedroom_0006/rgb_0186_disparity.npz
-NYUv2/test/bedroom_0006/rgb_0187_rgb_left.mp4,NYUv2/test/bedroom_0006/rgb_0187_disparity.npz
-NYUv2/test/bedroom_0006/rgb_0188_rgb_left.mp4,NYUv2/test/bedroom_0006/rgb_0188_disparity.npz
-NYUv2/test/bedroom_0007/rgb_0189_rgb_left.mp4,NYUv2/test/bedroom_0007/rgb_0189_disparity.npz
-NYUv2/test/bedroom_0007/rgb_0190_rgb_left.mp4,NYUv2/test/bedroom_0007/rgb_0190_disparity.npz
-NYUv2/test/bedroom_0007/rgb_0191_rgb_left.mp4,NYUv2/test/bedroom_0007/rgb_0191_disparity.npz
-NYUv2/test/bedroom_0007/rgb_0192_rgb_left.mp4,NYUv2/test/bedroom_0007/rgb_0192_disparity.npz
-NYUv2/test/bedroom_0007/rgb_0193_rgb_left.mp4,NYUv2/test/bedroom_0007/rgb_0193_disparity.npz
-NYUv2/test/kitchen_0002/rgb_0194_rgb_left.mp4,NYUv2/test/kitchen_0002/rgb_0194_disparity.npz
-NYUv2/test/kitchen_0002/rgb_0195_rgb_left.mp4,NYUv2/test/kitchen_0002/rgb_0195_disparity.npz
-NYUv2/test/kitchen_0002/rgb_0196_rgb_left.mp4,NYUv2/test/kitchen_0002/rgb_0196_disparity.npz
-NYUv2/test/kitchen_0002/rgb_0197_rgb_left.mp4,NYUv2/test/kitchen_0002/rgb_0197_disparity.npz
-NYUv2/test/kitchen_0002/rgb_0198_rgb_left.mp4,NYUv2/test/kitchen_0002/rgb_0198_disparity.npz
-NYUv2/test/kitchen_0002/rgb_0199_rgb_left.mp4,NYUv2/test/kitchen_0002/rgb_0199_disparity.npz
-NYUv2/test/kitchen_0002/rgb_0200_rgb_left.mp4,NYUv2/test/kitchen_0002/rgb_0200_disparity.npz
-NYUv2/test/kitchen_0002/rgb_0201_rgb_left.mp4,NYUv2/test/kitchen_0002/rgb_0201_disparity.npz
-NYUv2/test/kitchen_0002/rgb_0202_rgb_left.mp4,NYUv2/test/kitchen_0002/rgb_0202_disparity.npz
-NYUv2/test/living_room_0002/rgb_0207_rgb_left.mp4,NYUv2/test/living_room_0002/rgb_0207_disparity.npz
-NYUv2/test/living_room_0003/rgb_0208_rgb_left.mp4,NYUv2/test/living_room_0003/rgb_0208_disparity.npz
-NYUv2/test/living_room_0003/rgb_0209_rgb_left.mp4,NYUv2/test/living_room_0003/rgb_0209_disparity.npz
-NYUv2/test/living_room_0003/rgb_0210_rgb_left.mp4,NYUv2/test/living_room_0003/rgb_0210_disparity.npz
-NYUv2/test/living_room_0003/rgb_0211_rgb_left.mp4,NYUv2/test/living_room_0003/rgb_0211_disparity.npz
-NYUv2/test/living_room_0003/rgb_0212_rgb_left.mp4,NYUv2/test/living_room_0003/rgb_0212_disparity.npz
-NYUv2/test/bedroom_0022/rgb_0220_rgb_left.mp4,NYUv2/test/bedroom_0022/rgb_0220_disparity.npz
-NYUv2/test/bedroom_0024/rgb_0221_rgb_left.mp4,NYUv2/test/bedroom_0024/rgb_0221_disparity.npz
-NYUv2/test/bedroom_0024/rgb_0222_rgb_left.mp4,NYUv2/test/bedroom_0024/rgb_0222_disparity.npz
-NYUv2/test/kitchen_0015/rgb_0250_rgb_left.mp4,NYUv2/test/kitchen_0015/rgb_0250_disparity.npz
-NYUv2/test/living_room_0021/rgb_0264_rgb_left.mp4,NYUv2/test/living_room_0021/rgb_0264_disparity.npz
-NYUv2/test/office_0016/rgb_0271_rgb_left.mp4,NYUv2/test/office_0016/rgb_0271_disparity.npz
-NYUv2/test/office_0017/rgb_0272_rgb_left.mp4,NYUv2/test/office_0017/rgb_0272_disparity.npz
-NYUv2/test/study_room_0001/rgb_0273_rgb_left.mp4,NYUv2/test/study_room_0001/rgb_0273_disparity.npz
-NYUv2/test/study_room_0006/rgb_0279_rgb_left.mp4,NYUv2/test/study_room_0006/rgb_0279_disparity.npz
-NYUv2/test/bedroom_0131/rgb_0280_rgb_left.mp4,NYUv2/test/bedroom_0131/rgb_0280_disparity.npz
-NYUv2/test/bedroom_0131/rgb_0281_rgb_left.mp4,NYUv2/test/bedroom_0131/rgb_0281_disparity.npz
-NYUv2/test/bedroom_0131/rgb_0282_rgb_left.mp4,NYUv2/test/bedroom_0131/rgb_0282_disparity.npz
-NYUv2/test/bedroom_0131/rgb_0283_rgb_left.mp4,NYUv2/test/bedroom_0131/rgb_0283_disparity.npz
-NYUv2/test/classroom_0001/rgb_0284_rgb_left.mp4,NYUv2/test/classroom_0001/rgb_0284_disparity.npz
-NYUv2/test/classroom_0001/rgb_0285_rgb_left.mp4,NYUv2/test/classroom_0001/rgb_0285_disparity.npz
-NYUv2/test/classroom_0007/rgb_0296_rgb_left.mp4,NYUv2/test/classroom_0007/rgb_0296_disparity.npz
-NYUv2/test/classroom_0007/rgb_0297_rgb_left.mp4,NYUv2/test/classroom_0007/rgb_0297_disparity.npz
-NYUv2/test/classroom_0007/rgb_0298_rgb_left.mp4,NYUv2/test/classroom_0007/rgb_0298_disparity.npz
-NYUv2/test/classroom_0008/rgb_0299_rgb_left.mp4,NYUv2/test/classroom_0008/rgb_0299_disparity.npz
-NYUv2/test/classroom_0008/rgb_0300_rgb_left.mp4,NYUv2/test/classroom_0008/rgb_0300_disparity.npz
-NYUv2/test/classroom_0009/rgb_0301_rgb_left.mp4,NYUv2/test/classroom_0009/rgb_0301_disparity.npz
-NYUv2/test/classroom_0009/rgb_0302_rgb_left.mp4,NYUv2/test/classroom_0009/rgb_0302_disparity.npz
-NYUv2/test/classroom_0014/rgb_0310_rgb_left.mp4,NYUv2/test/classroom_0014/rgb_0310_disparity.npz
-NYUv2/test/classroom_0014/rgb_0311_rgb_left.mp4,NYUv2/test/classroom_0014/rgb_0311_disparity.npz
-NYUv2/test/classroom_0015/rgb_0312_rgb_left.mp4,NYUv2/test/classroom_0015/rgb_0312_disparity.npz
-NYUv2/test/classroom_0017/rgb_0315_rgb_left.mp4,NYUv2/test/classroom_0017/rgb_0315_disparity.npz
-NYUv2/test/classroom_0017/rgb_0316_rgb_left.mp4,NYUv2/test/classroom_0017/rgb_0316_disparity.npz
-NYUv2/test/classroom_0017/rgb_0317_rgb_left.mp4,NYUv2/test/classroom_0017/rgb_0317_disparity.npz
-NYUv2/test/classroom_0023/rgb_0325_rgb_left.mp4,NYUv2/test/classroom_0023/rgb_0325_disparity.npz
-NYUv2/test/classroom_0023/rgb_0326_rgb_left.mp4,NYUv2/test/classroom_0023/rgb_0326_disparity.npz
-NYUv2/test/classroom_0023/rgb_0327_rgb_left.mp4,NYUv2/test/classroom_0023/rgb_0327_disparity.npz
-NYUv2/test/classroom_0023/rgb_0328_rgb_left.mp4,NYUv2/test/classroom_0023/rgb_0328_disparity.npz
-NYUv2/test/classroom_0024/rgb_0329_rgb_left.mp4,NYUv2/test/classroom_0024/rgb_0329_disparity.npz
-NYUv2/test/classroom_0024/rgb_0330_rgb_left.mp4,NYUv2/test/classroom_0024/rgb_0330_disparity.npz
-NYUv2/test/classroom_0026/rgb_0331_rgb_left.mp4,NYUv2/test/classroom_0026/rgb_0331_disparity.npz
-NYUv2/test/classroom_0026/rgb_0332_rgb_left.mp4,NYUv2/test/classroom_0026/rgb_0332_disparity.npz
-NYUv2/test/computer_lab_0001/rgb_0333_rgb_left.mp4,NYUv2/test/computer_lab_0001/rgb_0333_disparity.npz
-NYUv2/test/computer_lab_0001/rgb_0334_rgb_left.mp4,NYUv2/test/computer_lab_0001/rgb_0334_disparity.npz
-NYUv2/test/computer_lab_0001/rgb_0335_rgb_left.mp4,NYUv2/test/computer_lab_0001/rgb_0335_disparity.npz
-NYUv2/test/foyer_0001/rgb_0351_rgb_left.mp4,NYUv2/test/foyer_0001/rgb_0351_disparity.npz
-NYUv2/test/foyer_0001/rgb_0352_rgb_left.mp4,NYUv2/test/foyer_0001/rgb_0352_disparity.npz
-NYUv2/test/home_office_0001/rgb_0355_rgb_left.mp4,NYUv2/test/home_office_0001/rgb_0355_disparity.npz
-NYUv2/test/home_office_0001/rgb_0356_rgb_left.mp4,NYUv2/test/home_office_0001/rgb_0356_disparity.npz
-NYUv2/test/home_office_0001/rgb_0357_rgb_left.mp4,NYUv2/test/home_office_0001/rgb_0357_disparity.npz
-NYUv2/test/home_office_0001/rgb_0358_rgb_left.mp4,NYUv2/test/home_office_0001/rgb_0358_disparity.npz
-NYUv2/test/home_office_0002/rgb_0359_rgb_left.mp4,NYUv2/test/home_office_0002/rgb_0359_disparity.npz
-NYUv2/test/home_office_0002/rgb_0360_rgb_left.mp4,NYUv2/test/home_office_0002/rgb_0360_disparity.npz
-NYUv2/test/home_office_0002/rgb_0361_rgb_left.mp4,NYUv2/test/home_office_0002/rgb_0361_disparity.npz
-NYUv2/test/home_office_0002/rgb_0362_rgb_left.mp4,NYUv2/test/home_office_0002/rgb_0362_disparity.npz
-NYUv2/test/home_office_0003/rgb_0363_rgb_left.mp4,NYUv2/test/home_office_0003/rgb_0363_disparity.npz
-NYUv2/test/home_office_0003/rgb_0364_rgb_left.mp4,NYUv2/test/home_office_0003/rgb_0364_disparity.npz
-NYUv2/test/home_office_0009/rgb_0384_rgb_left.mp4,NYUv2/test/home_office_0009/rgb_0384_disparity.npz
-NYUv2/test/home_office_0009/rgb_0385_rgb_left.mp4,NYUv2/test/home_office_0009/rgb_0385_disparity.npz
-NYUv2/test/home_office_0009/rgb_0386_rgb_left.mp4,NYUv2/test/home_office_0009/rgb_0386_disparity.npz
-NYUv2/test/home_office_0010/rgb_0387_rgb_left.mp4,NYUv2/test/home_office_0010/rgb_0387_disparity.npz
-NYUv2/test/home_office_0010/rgb_0388_rgb_left.mp4,NYUv2/test/home_office_0010/rgb_0388_disparity.npz
-NYUv2/test/home_office_0010/rgb_0389_rgb_left.mp4,NYUv2/test/home_office_0010/rgb_0389_disparity.npz
-NYUv2/test/home_office_0010/rgb_0390_rgb_left.mp4,NYUv2/test/home_office_0010/rgb_0390_disparity.npz
-NYUv2/test/home_office_0012/rgb_0395_rgb_left.mp4,NYUv2/test/home_office_0012/rgb_0395_disparity.npz
-NYUv2/test/home_office_0012/rgb_0396_rgb_left.mp4,NYUv2/test/home_office_0012/rgb_0396_disparity.npz
-NYUv2/test/home_office_0012/rgb_0397_rgb_left.mp4,NYUv2/test/home_office_0012/rgb_0397_disparity.npz
-NYUv2/test/office_kitchen_0002/rgb_0411_rgb_left.mp4,NYUv2/test/office_kitchen_0002/rgb_0411_disparity.npz
-NYUv2/test/office_kitchen_0002/rgb_0412_rgb_left.mp4,NYUv2/test/office_kitchen_0002/rgb_0412_disparity.npz
-NYUv2/test/office_kitchen_0002/rgb_0413_rgb_left.mp4,NYUv2/test/office_kitchen_0002/rgb_0413_disparity.npz
-NYUv2/test/office_kitchen_0002/rgb_0414_rgb_left.mp4,NYUv2/test/office_kitchen_0002/rgb_0414_disparity.npz
-NYUv2/test/playroom_0005/rgb_0430_rgb_left.mp4,NYUv2/test/playroom_0005/rgb_0430_disparity.npz
-NYUv2/test/playroom_0005/rgb_0431_rgb_left.mp4,NYUv2/test/playroom_0005/rgb_0431_disparity.npz
-NYUv2/test/playroom_0005/rgb_0432_rgb_left.mp4,NYUv2/test/playroom_0005/rgb_0432_disparity.npz
-NYUv2/test/playroom_0005/rgb_0433_rgb_left.mp4,NYUv2/test/playroom_0005/rgb_0433_disparity.npz
-NYUv2/test/playroom_0005/rgb_0434_rgb_left.mp4,NYUv2/test/playroom_0005/rgb_0434_disparity.npz
-NYUv2/test/playroom_0005/rgb_0435_rgb_left.mp4,NYUv2/test/playroom_0005/rgb_0435_disparity.npz
-NYUv2/test/playroom_0007/rgb_0441_rgb_left.mp4,NYUv2/test/playroom_0007/rgb_0441_disparity.npz
-NYUv2/test/playroom_0007/rgb_0442_rgb_left.mp4,NYUv2/test/playroom_0007/rgb_0442_disparity.npz
-NYUv2/test/playroom_0007/rgb_0443_rgb_left.mp4,NYUv2/test/playroom_0007/rgb_0443_disparity.npz
-NYUv2/test/playroom_0008/rgb_0444_rgb_left.mp4,NYUv2/test/playroom_0008/rgb_0444_disparity.npz
-NYUv2/test/playroom_0008/rgb_0445_rgb_left.mp4,NYUv2/test/playroom_0008/rgb_0445_disparity.npz
-NYUv2/test/playroom_0008/rgb_0446_rgb_left.mp4,NYUv2/test/playroom_0008/rgb_0446_disparity.npz
-NYUv2/test/playroom_0008/rgb_0447_rgb_left.mp4,NYUv2/test/playroom_0008/rgb_0447_disparity.npz
-NYUv2/test/playroom_0008/rgb_0448_rgb_left.mp4,NYUv2/test/playroom_0008/rgb_0448_disparity.npz
-NYUv2/test/reception_room_0003/rgb_0462_rgb_left.mp4,NYUv2/test/reception_room_0003/rgb_0462_disparity.npz
-NYUv2/test/reception_room_0003/rgb_0463_rgb_left.mp4,NYUv2/test/reception_room_0003/rgb_0463_disparity.npz
-NYUv2/test/reception_room_0003/rgb_0464_rgb_left.mp4,NYUv2/test/reception_room_0003/rgb_0464_disparity.npz
-NYUv2/test/reception_room_0003/rgb_0465_rgb_left.mp4,NYUv2/test/reception_room_0003/rgb_0465_disparity.npz
-NYUv2/test/reception_room_0003/rgb_0466_rgb_left.mp4,NYUv2/test/reception_room_0003/rgb_0466_disparity.npz
-NYUv2/test/study_0001/rgb_0469_rgb_left.mp4,NYUv2/test/study_0001/rgb_0469_disparity.npz
-NYUv2/test/study_0001/rgb_0470_rgb_left.mp4,NYUv2/test/study_0001/rgb_0470_disparity.npz
-NYUv2/test/study_0001/rgb_0471_rgb_left.mp4,NYUv2/test/study_0001/rgb_0471_disparity.npz
-NYUv2/test/study_0001/rgb_0472_rgb_left.mp4,NYUv2/test/study_0001/rgb_0472_disparity.npz
-NYUv2/test/study_0001/rgb_0473_rgb_left.mp4,NYUv2/test/study_0001/rgb_0473_disparity.npz
-NYUv2/test/study_0002/rgb_0474_rgb_left.mp4,NYUv2/test/study_0002/rgb_0474_disparity.npz
-NYUv2/test/study_0002/rgb_0475_rgb_left.mp4,NYUv2/test/study_0002/rgb_0475_disparity.npz
-NYUv2/test/study_0002/rgb_0476_rgb_left.mp4,NYUv2/test/study_0002/rgb_0476_disparity.npz
-NYUv2/test/study_0002/rgb_0477_rgb_left.mp4,NYUv2/test/study_0002/rgb_0477_disparity.npz
-NYUv2/test/bathroom_0058/rgb_0508_rgb_left.mp4,NYUv2/test/bathroom_0058/rgb_0508_disparity.npz
-NYUv2/test/bathroom_0058/rgb_0509_rgb_left.mp4,NYUv2/test/bathroom_0058/rgb_0509_disparity.npz
-NYUv2/test/bathroom_0058/rgb_0510_rgb_left.mp4,NYUv2/test/bathroom_0058/rgb_0510_disparity.npz
-NYUv2/test/bathroom_0060/rgb_0511_rgb_left.mp4,NYUv2/test/bathroom_0060/rgb_0511_disparity.npz
-NYUv2/test/bathroom_0060/rgb_0512_rgb_left.mp4,NYUv2/test/bathroom_0060/rgb_0512_disparity.npz
-NYUv2/test/bathroom_0060/rgb_0513_rgb_left.mp4,NYUv2/test/bathroom_0060/rgb_0513_disparity.npz
-NYUv2/test/bedroom_0133/rgb_0515_rgb_left.mp4,NYUv2/test/bedroom_0133/rgb_0515_disparity.npz
-NYUv2/test/bedroom_0133/rgb_0516_rgb_left.mp4,NYUv2/test/bedroom_0133/rgb_0516_disparity.npz
-NYUv2/test/bedroom_0133/rgb_0517_rgb_left.mp4,NYUv2/test/bedroom_0133/rgb_0517_disparity.npz
-NYUv2/test/bedroom_0133/rgb_0518_rgb_left.mp4,NYUv2/test/bedroom_0133/rgb_0518_disparity.npz
-NYUv2/test/bedroom_0133/rgb_0519_rgb_left.mp4,NYUv2/test/bedroom_0133/rgb_0519_disparity.npz
-NYUv2/test/bedroom_0134/rgb_0520_rgb_left.mp4,NYUv2/test/bedroom_0134/rgb_0520_disparity.npz
-NYUv2/test/bedroom_0134/rgb_0521_rgb_left.mp4,NYUv2/test/bedroom_0134/rgb_0521_disparity.npz
-NYUv2/test/bedroom_0134/rgb_0522_rgb_left.mp4,NYUv2/test/bedroom_0134/rgb_0522_disparity.npz
-NYUv2/test/bedroom_0135/rgb_0523_rgb_left.mp4,NYUv2/test/bedroom_0135/rgb_0523_disparity.npz
-NYUv2/test/bedroom_0135/rgb_0524_rgb_left.mp4,NYUv2/test/bedroom_0135/rgb_0524_disparity.npz
-NYUv2/test/bedroom_0135/rgb_0525_rgb_left.mp4,NYUv2/test/bedroom_0135/rgb_0525_disparity.npz
-NYUv2/test/bedroom_0135/rgb_0526_rgb_left.mp4,NYUv2/test/bedroom_0135/rgb_0526_disparity.npz
-NYUv2/test/bedroom_0137/rgb_0531_rgb_left.mp4,NYUv2/test/bedroom_0137/rgb_0531_disparity.npz
-NYUv2/test/bedroom_0137/rgb_0532_rgb_left.mp4,NYUv2/test/bedroom_0137/rgb_0532_disparity.npz
-NYUv2/test/bedroom_0137/rgb_0533_rgb_left.mp4,NYUv2/test/bedroom_0137/rgb_0533_disparity.npz
-NYUv2/test/bedroom_0139/rgb_0537_rgb_left.mp4,NYUv2/test/bedroom_0139/rgb_0537_disparity.npz
-NYUv2/test/bedroom_0139/rgb_0538_rgb_left.mp4,NYUv2/test/bedroom_0139/rgb_0538_disparity.npz
-NYUv2/test/bedroom_0139/rgb_0539_rgb_left.mp4,NYUv2/test/bedroom_0139/rgb_0539_disparity.npz
-NYUv2/test/dining_room_0038/rgb_0549_rgb_left.mp4,NYUv2/test/dining_room_0038/rgb_0549_disparity.npz
-NYUv2/test/dining_room_0038/rgb_0550_rgb_left.mp4,NYUv2/test/dining_room_0038/rgb_0550_disparity.npz
-NYUv2/test/dining_room_0038/rgb_0551_rgb_left.mp4,NYUv2/test/dining_room_0038/rgb_0551_disparity.npz
-NYUv2/test/home_office_0014/rgb_0555_rgb_left.mp4,NYUv2/test/home_office_0014/rgb_0555_disparity.npz
-NYUv2/test/home_office_0014/rgb_0556_rgb_left.mp4,NYUv2/test/home_office_0014/rgb_0556_disparity.npz
-NYUv2/test/home_office_0014/rgb_0557_rgb_left.mp4,NYUv2/test/home_office_0014/rgb_0557_disparity.npz
-NYUv2/test/home_office_0014/rgb_0558_rgb_left.mp4,NYUv2/test/home_office_0014/rgb_0558_disparity.npz
-NYUv2/test/kitchen_0055/rgb_0559_rgb_left.mp4,NYUv2/test/kitchen_0055/rgb_0559_disparity.npz
-NYUv2/test/kitchen_0055/rgb_0560_rgb_left.mp4,NYUv2/test/kitchen_0055/rgb_0560_disparity.npz
-NYUv2/test/kitchen_0056/rgb_0561_rgb_left.mp4,NYUv2/test/kitchen_0056/rgb_0561_disparity.npz
-NYUv2/test/kitchen_0056/rgb_0562_rgb_left.mp4,NYUv2/test/kitchen_0056/rgb_0562_disparity.npz
-NYUv2/test/kitchen_0056/rgb_0563_rgb_left.mp4,NYUv2/test/kitchen_0056/rgb_0563_disparity.npz
-NYUv2/test/kitchen_0056/rgb_0564_rgb_left.mp4,NYUv2/test/kitchen_0056/rgb_0564_disparity.npz
-NYUv2/test/kitchen_0057/rgb_0565_rgb_left.mp4,NYUv2/test/kitchen_0057/rgb_0565_disparity.npz
-NYUv2/test/kitchen_0057/rgb_0566_rgb_left.mp4,NYUv2/test/kitchen_0057/rgb_0566_disparity.npz
-NYUv2/test/kitchen_0057/rgb_0567_rgb_left.mp4,NYUv2/test/kitchen_0057/rgb_0567_disparity.npz
-NYUv2/test/kitchen_0057/rgb_0568_rgb_left.mp4,NYUv2/test/kitchen_0057/rgb_0568_disparity.npz
-NYUv2/test/kitchen_0058/rgb_0569_rgb_left.mp4,NYUv2/test/kitchen_0058/rgb_0569_disparity.npz
-NYUv2/test/kitchen_0058/rgb_0570_rgb_left.mp4,NYUv2/test/kitchen_0058/rgb_0570_disparity.npz
-NYUv2/test/kitchen_0058/rgb_0571_rgb_left.mp4,NYUv2/test/kitchen_0058/rgb_0571_disparity.npz
-NYUv2/test/living_room_0081/rgb_0579_rgb_left.mp4,NYUv2/test/living_room_0081/rgb_0579_disparity.npz
-NYUv2/test/living_room_0081/rgb_0580_rgb_left.mp4,NYUv2/test/living_room_0081/rgb_0580_disparity.npz
-NYUv2/test/living_room_0081/rgb_0581_rgb_left.mp4,NYUv2/test/living_room_0081/rgb_0581_disparity.npz
-NYUv2/test/living_room_0081/rgb_0582_rgb_left.mp4,NYUv2/test/living_room_0081/rgb_0582_disparity.npz
-NYUv2/test/living_room_0081/rgb_0583_rgb_left.mp4,NYUv2/test/living_room_0081/rgb_0583_disparity.npz
-NYUv2/test/living_room_0084/rgb_0591_rgb_left.mp4,NYUv2/test/living_room_0084/rgb_0591_disparity.npz
-NYUv2/test/living_room_0084/rgb_0592_rgb_left.mp4,NYUv2/test/living_room_0084/rgb_0592_disparity.npz
-NYUv2/test/living_room_0084/rgb_0593_rgb_left.mp4,NYUv2/test/living_room_0084/rgb_0593_disparity.npz
-NYUv2/test/living_room_0084/rgb_0594_rgb_left.mp4,NYUv2/test/living_room_0084/rgb_0594_disparity.npz
-NYUv2/test/living_room_0087/rgb_0603_rgb_left.mp4,NYUv2/test/living_room_0087/rgb_0603_disparity.npz
-NYUv2/test/living_room_0087/rgb_0604_rgb_left.mp4,NYUv2/test/living_room_0087/rgb_0604_disparity.npz
-NYUv2/test/living_room_0087/rgb_0605_rgb_left.mp4,NYUv2/test/living_room_0087/rgb_0605_disparity.npz
-NYUv2/test/living_room_0087/rgb_0606_rgb_left.mp4,NYUv2/test/living_room_0087/rgb_0606_disparity.npz
-NYUv2/test/living_room_0087/rgb_0607_rgb_left.mp4,NYUv2/test/living_room_0087/rgb_0607_disparity.npz
-NYUv2/test/office_0020/rgb_0612_rgb_left.mp4,NYUv2/test/office_0020/rgb_0612_disparity.npz
-NYUv2/test/office_0020/rgb_0613_rgb_left.mp4,NYUv2/test/office_0020/rgb_0613_disparity.npz
-NYUv2/test/office_0022/rgb_0617_rgb_left.mp4,NYUv2/test/office_0022/rgb_0617_disparity.npz
-NYUv2/test/office_0022/rgb_0618_rgb_left.mp4,NYUv2/test/office_0022/rgb_0618_disparity.npz
-NYUv2/test/office_0022/rgb_0619_rgb_left.mp4,NYUv2/test/office_0022/rgb_0619_disparity.npz
-NYUv2/test/office_0022/rgb_0620_rgb_left.mp4,NYUv2/test/office_0022/rgb_0620_disparity.npz
-NYUv2/test/office_0022/rgb_0621_rgb_left.mp4,NYUv2/test/office_0022/rgb_0621_disparity.npz
-NYUv2/test/office_0027/rgb_0633_rgb_left.mp4,NYUv2/test/office_0027/rgb_0633_disparity.npz
-NYUv2/test/office_0027/rgb_0634_rgb_left.mp4,NYUv2/test/office_0027/rgb_0634_disparity.npz
-NYUv2/test/office_0027/rgb_0635_rgb_left.mp4,NYUv2/test/office_0027/rgb_0635_disparity.npz
-NYUv2/test/office_0027/rgb_0636_rgb_left.mp4,NYUv2/test/office_0027/rgb_0636_disparity.npz
-NYUv2/test/office_0027/rgb_0637_rgb_left.mp4,NYUv2/test/office_0027/rgb_0637_disparity.npz
-NYUv2/test/office_0027/rgb_0638_rgb_left.mp4,NYUv2/test/office_0027/rgb_0638_disparity.npz
-NYUv2/test/study_0007/rgb_0644_rgb_left.mp4,NYUv2/test/study_0007/rgb_0644_disparity.npz
-NYUv2/test/study_0007/rgb_0645_rgb_left.mp4,NYUv2/test/study_0007/rgb_0645_disparity.npz
-NYUv2/test/bathroom_0008/rgb_0650_rgb_left.mp4,NYUv2/test/bathroom_0008/rgb_0650_disparity.npz
-NYUv2/test/bathroom_0009/rgb_0651_rgb_left.mp4,NYUv2/test/bathroom_0009/rgb_0651_disparity.npz
-NYUv2/test/bathroom_0012/rgb_0656_rgb_left.mp4,NYUv2/test/bathroom_0012/rgb_0656_disparity.npz
-NYUv2/test/bathroom_0012/rgb_0657_rgb_left.mp4,NYUv2/test/bathroom_0012/rgb_0657_disparity.npz
-NYUv2/test/bathroom_0012/rgb_0658_rgb_left.mp4,NYUv2/test/bathroom_0012/rgb_0658_disparity.npz
-NYUv2/test/bathroom_0015/rgb_0663_rgb_left.mp4,NYUv2/test/bathroom_0015/rgb_0663_disparity.npz
-NYUv2/test/bathroom_0015/rgb_0664_rgb_left.mp4,NYUv2/test/bathroom_0015/rgb_0664_disparity.npz
-NYUv2/test/bathroom_0017/rgb_0668_rgb_left.mp4,NYUv2/test/bathroom_0017/rgb_0668_disparity.npz
-NYUv2/test/bathroom_0017/rgb_0669_rgb_left.mp4,NYUv2/test/bathroom_0017/rgb_0669_disparity.npz
-NYUv2/test/bathroom_0017/rgb_0670_rgb_left.mp4,NYUv2/test/bathroom_0017/rgb_0670_disparity.npz
-NYUv2/test/bathroom_0018/rgb_0671_rgb_left.mp4,NYUv2/test/bathroom_0018/rgb_0671_disparity.npz
-NYUv2/test/bathroom_0018/rgb_0672_rgb_left.mp4,NYUv2/test/bathroom_0018/rgb_0672_disparity.npz
-NYUv2/test/bathroom_0018/rgb_0673_rgb_left.mp4,NYUv2/test/bathroom_0018/rgb_0673_disparity.npz
-NYUv2/test/bathroom_0020/rgb_0676_rgb_left.mp4,NYUv2/test/bathroom_0020/rgb_0676_disparity.npz
-NYUv2/test/bathroom_0020/rgb_0677_rgb_left.mp4,NYUv2/test/bathroom_0020/rgb_0677_disparity.npz
-NYUv2/test/bathroom_0021/rgb_0678_rgb_left.mp4,NYUv2/test/bathroom_0021/rgb_0678_disparity.npz
-NYUv2/test/bathroom_0021/rgb_0679_rgb_left.mp4,NYUv2/test/bathroom_0021/rgb_0679_disparity.npz
-NYUv2/test/bathroom_0022/rgb_0680_rgb_left.mp4,NYUv2/test/bathroom_0022/rgb_0680_disparity.npz
-NYUv2/test/bathroom_0022/rgb_0681_rgb_left.mp4,NYUv2/test/bathroom_0022/rgb_0681_disparity.npz
-NYUv2/test/bathroom_0025/rgb_0686_rgb_left.mp4,NYUv2/test/bathroom_0025/rgb_0686_disparity.npz
-NYUv2/test/bathroom_0025/rgb_0687_rgb_left.mp4,NYUv2/test/bathroom_0025/rgb_0687_disparity.npz
-NYUv2/test/bathroom_0026/rgb_0688_rgb_left.mp4,NYUv2/test/bathroom_0026/rgb_0688_disparity.npz
-NYUv2/test/bathroom_0026/rgb_0689_rgb_left.mp4,NYUv2/test/bathroom_0026/rgb_0689_disparity.npz
-NYUv2/test/bathroom_0026/rgb_0690_rgb_left.mp4,NYUv2/test/bathroom_0026/rgb_0690_disparity.npz
-NYUv2/test/bathroom_0029/rgb_0693_rgb_left.mp4,NYUv2/test/bathroom_0029/rgb_0693_disparity.npz
-NYUv2/test/bathroom_0029/rgb_0694_rgb_left.mp4,NYUv2/test/bathroom_0029/rgb_0694_disparity.npz
-NYUv2/test/bathroom_0031/rgb_0697_rgb_left.mp4,NYUv2/test/bathroom_0031/rgb_0697_disparity.npz
-NYUv2/test/bathroom_0031/rgb_0698_rgb_left.mp4,NYUv2/test/bathroom_0031/rgb_0698_disparity.npz
-NYUv2/test/bathroom_0031/rgb_0699_rgb_left.mp4,NYUv2/test/bathroom_0031/rgb_0699_disparity.npz
-NYUv2/test/bathroom_0036/rgb_0706_rgb_left.mp4,NYUv2/test/bathroom_0036/rgb_0706_disparity.npz
-NYUv2/test/bathroom_0036/rgb_0707_rgb_left.mp4,NYUv2/test/bathroom_0036/rgb_0707_disparity.npz
-NYUv2/test/bathroom_0036/rgb_0708_rgb_left.mp4,NYUv2/test/bathroom_0036/rgb_0708_disparity.npz
-NYUv2/test/bathroom_0037/rgb_0709_rgb_left.mp4,NYUv2/test/bathroom_0037/rgb_0709_disparity.npz
-NYUv2/test/bathroom_0037/rgb_0710_rgb_left.mp4,NYUv2/test/bathroom_0037/rgb_0710_disparity.npz
-NYUv2/test/bathroom_0038/rgb_0711_rgb_left.mp4,NYUv2/test/bathroom_0038/rgb_0711_disparity.npz
-NYUv2/test/bathroom_0038/rgb_0712_rgb_left.mp4,NYUv2/test/bathroom_0038/rgb_0712_disparity.npz
-NYUv2/test/bathroom_0038/rgb_0713_rgb_left.mp4,NYUv2/test/bathroom_0038/rgb_0713_disparity.npz
-NYUv2/test/bathroom_0040/rgb_0717_rgb_left.mp4,NYUv2/test/bathroom_0040/rgb_0717_disparity.npz
-NYUv2/test/bathroom_0040/rgb_0718_rgb_left.mp4,NYUv2/test/bathroom_0040/rgb_0718_disparity.npz
-NYUv2/test/bathroom_0043/rgb_0724_rgb_left.mp4,NYUv2/test/bathroom_0043/rgb_0724_disparity.npz
-NYUv2/test/bathroom_0043/rgb_0725_rgb_left.mp4,NYUv2/test/bathroom_0043/rgb_0725_disparity.npz
-NYUv2/test/bathroom_0043/rgb_0726_rgb_left.mp4,NYUv2/test/bathroom_0043/rgb_0726_disparity.npz
-NYUv2/test/bathroom_0044/rgb_0727_rgb_left.mp4,NYUv2/test/bathroom_0044/rgb_0727_disparity.npz
-NYUv2/test/bathroom_0044/rgb_0728_rgb_left.mp4,NYUv2/test/bathroom_0044/rgb_0728_disparity.npz
-NYUv2/test/bathroom_0046/rgb_0731_rgb_left.mp4,NYUv2/test/bathroom_0046/rgb_0731_disparity.npz
-NYUv2/test/bathroom_0046/rgb_0732_rgb_left.mp4,NYUv2/test/bathroom_0046/rgb_0732_disparity.npz
-NYUv2/test/bathroom_0047/rgb_0733_rgb_left.mp4,NYUv2/test/bathroom_0047/rgb_0733_disparity.npz
-NYUv2/test/bathroom_0047/rgb_0734_rgb_left.mp4,NYUv2/test/bathroom_0047/rgb_0734_disparity.npz
-NYUv2/test/bathroom_0052/rgb_0743_rgb_left.mp4,NYUv2/test/bathroom_0052/rgb_0743_disparity.npz
-NYUv2/test/bathroom_0052/rgb_0744_rgb_left.mp4,NYUv2/test/bathroom_0052/rgb_0744_disparity.npz
-NYUv2/test/kitchen_0021/rgb_0759_rgb_left.mp4,NYUv2/test/kitchen_0021/rgb_0759_disparity.npz
-NYUv2/test/kitchen_0021/rgb_0760_rgb_left.mp4,NYUv2/test/kitchen_0021/rgb_0760_disparity.npz
-NYUv2/test/kitchen_0021/rgb_0761_rgb_left.mp4,NYUv2/test/kitchen_0021/rgb_0761_disparity.npz
-NYUv2/test/kitchen_0022/rgb_0762_rgb_left.mp4,NYUv2/test/kitchen_0022/rgb_0762_disparity.npz
-NYUv2/test/kitchen_0022/rgb_0763_rgb_left.mp4,NYUv2/test/kitchen_0022/rgb_0763_disparity.npz
-NYUv2/test/kitchen_0022/rgb_0764_rgb_left.mp4,NYUv2/test/kitchen_0022/rgb_0764_disparity.npz
-NYUv2/test/kitchen_0022/rgb_0765_rgb_left.mp4,NYUv2/test/kitchen_0022/rgb_0765_disparity.npz
-NYUv2/test/kitchen_0022/rgb_0766_rgb_left.mp4,NYUv2/test/kitchen_0022/rgb_0766_disparity.npz
-NYUv2/test/kitchen_0023/rgb_0767_rgb_left.mp4,NYUv2/test/kitchen_0023/rgb_0767_disparity.npz
-NYUv2/test/kitchen_0023/rgb_0768_rgb_left.mp4,NYUv2/test/kitchen_0023/rgb_0768_disparity.npz
-NYUv2/test/kitchen_0023/rgb_0769_rgb_left.mp4,NYUv2/test/kitchen_0023/rgb_0769_disparity.npz
-NYUv2/test/kitchen_0023/rgb_0770_rgb_left.mp4,NYUv2/test/kitchen_0023/rgb_0770_disparity.npz
-NYUv2/test/kitchen_0023/rgb_0771_rgb_left.mp4,NYUv2/test/kitchen_0023/rgb_0771_disparity.npz
-NYUv2/test/kitchen_0024/rgb_0772_rgb_left.mp4,NYUv2/test/kitchen_0024/rgb_0772_disparity.npz
-NYUv2/test/kitchen_0024/rgb_0773_rgb_left.mp4,NYUv2/test/kitchen_0024/rgb_0773_disparity.npz
-NYUv2/test/kitchen_0024/rgb_0774_rgb_left.mp4,NYUv2/test/kitchen_0024/rgb_0774_disparity.npz
-NYUv2/test/kitchen_0024/rgb_0775_rgb_left.mp4,NYUv2/test/kitchen_0024/rgb_0775_disparity.npz
-NYUv2/test/kitchen_0024/rgb_0776_rgb_left.mp4,NYUv2/test/kitchen_0024/rgb_0776_disparity.npz
-NYUv2/test/kitchen_0025/rgb_0777_rgb_left.mp4,NYUv2/test/kitchen_0025/rgb_0777_disparity.npz
-NYUv2/test/kitchen_0025/rgb_0778_rgb_left.mp4,NYUv2/test/kitchen_0025/rgb_0778_disparity.npz
-NYUv2/test/kitchen_0025/rgb_0779_rgb_left.mp4,NYUv2/test/kitchen_0025/rgb_0779_disparity.npz
-NYUv2/test/kitchen_0026/rgb_0780_rgb_left.mp4,NYUv2/test/kitchen_0026/rgb_0780_disparity.npz
-NYUv2/test/kitchen_0026/rgb_0781_rgb_left.mp4,NYUv2/test/kitchen_0026/rgb_0781_disparity.npz
-NYUv2/test/kitchen_0026/rgb_0782_rgb_left.mp4,NYUv2/test/kitchen_0026/rgb_0782_disparity.npz
-NYUv2/test/kitchen_0027/rgb_0783_rgb_left.mp4,NYUv2/test/kitchen_0027/rgb_0783_disparity.npz
-NYUv2/test/kitchen_0027/rgb_0784_rgb_left.mp4,NYUv2/test/kitchen_0027/rgb_0784_disparity.npz
-NYUv2/test/kitchen_0027/rgb_0785_rgb_left.mp4,NYUv2/test/kitchen_0027/rgb_0785_disparity.npz
-NYUv2/test/kitchen_0027/rgb_0786_rgb_left.mp4,NYUv2/test/kitchen_0027/rgb_0786_disparity.npz
-NYUv2/test/kitchen_0027/rgb_0787_rgb_left.mp4,NYUv2/test/kitchen_0027/rgb_0787_disparity.npz
-NYUv2/test/kitchen_0030/rgb_0800_rgb_left.mp4,NYUv2/test/kitchen_0030/rgb_0800_disparity.npz
-NYUv2/test/kitchen_0030/rgb_0801_rgb_left.mp4,NYUv2/test/kitchen_0030/rgb_0801_disparity.npz
-NYUv2/test/kitchen_0030/rgb_0802_rgb_left.mp4,NYUv2/test/kitchen_0030/rgb_0802_disparity.npz
-NYUv2/test/kitchen_0030/rgb_0803_rgb_left.mp4,NYUv2/test/kitchen_0030/rgb_0803_disparity.npz
-NYUv2/test/kitchen_0030/rgb_0804_rgb_left.mp4,NYUv2/test/kitchen_0030/rgb_0804_disparity.npz
-NYUv2/test/kitchen_0032/rgb_0810_rgb_left.mp4,NYUv2/test/kitchen_0032/rgb_0810_disparity.npz
-NYUv2/test/kitchen_0032/rgb_0811_rgb_left.mp4,NYUv2/test/kitchen_0032/rgb_0811_disparity.npz
-NYUv2/test/kitchen_0032/rgb_0812_rgb_left.mp4,NYUv2/test/kitchen_0032/rgb_0812_disparity.npz
-NYUv2/test/kitchen_0032/rgb_0813_rgb_left.mp4,NYUv2/test/kitchen_0032/rgb_0813_disparity.npz
-NYUv2/test/kitchen_0032/rgb_0814_rgb_left.mp4,NYUv2/test/kitchen_0032/rgb_0814_disparity.npz
-NYUv2/test/kitchen_0034/rgb_0821_rgb_left.mp4,NYUv2/test/kitchen_0034/rgb_0821_disparity.npz
-NYUv2/test/kitchen_0034/rgb_0822_rgb_left.mp4,NYUv2/test/kitchen_0034/rgb_0822_disparity.npz
-NYUv2/test/kitchen_0034/rgb_0823_rgb_left.mp4,NYUv2/test/kitchen_0034/rgb_0823_disparity.npz
-NYUv2/test/kitchen_0038/rgb_0833_rgb_left.mp4,NYUv2/test/kitchen_0038/rgb_0833_disparity.npz
-NYUv2/test/kitchen_0038/rgb_0834_rgb_left.mp4,NYUv2/test/kitchen_0038/rgb_0834_disparity.npz
-NYUv2/test/kitchen_0038/rgb_0835_rgb_left.mp4,NYUv2/test/kitchen_0038/rgb_0835_disparity.npz
-NYUv2/test/kitchen_0038/rgb_0836_rgb_left.mp4,NYUv2/test/kitchen_0038/rgb_0836_disparity.npz
-NYUv2/test/kitchen_0039/rgb_0837_rgb_left.mp4,NYUv2/test/kitchen_0039/rgb_0837_disparity.npz
-NYUv2/test/kitchen_0039/rgb_0838_rgb_left.mp4,NYUv2/test/kitchen_0039/rgb_0838_disparity.npz
-NYUv2/test/kitchen_0039/rgb_0839_rgb_left.mp4,NYUv2/test/kitchen_0039/rgb_0839_disparity.npz
-NYUv2/test/kitchen_0039/rgb_0840_rgb_left.mp4,NYUv2/test/kitchen_0039/rgb_0840_disparity.npz
-NYUv2/test/kitchen_0039/rgb_0841_rgb_left.mp4,NYUv2/test/kitchen_0039/rgb_0841_disparity.npz
-NYUv2/test/kitchen_0039/rgb_0842_rgb_left.mp4,NYUv2/test/kitchen_0039/rgb_0842_disparity.npz
-NYUv2/test/kitchen_0040/rgb_0843_rgb_left.mp4,NYUv2/test/kitchen_0040/rgb_0843_disparity.npz
-NYUv2/test/kitchen_0040/rgb_0844_rgb_left.mp4,NYUv2/test/kitchen_0040/rgb_0844_disparity.npz
-NYUv2/test/kitchen_0040/rgb_0845_rgb_left.mp4,NYUv2/test/kitchen_0040/rgb_0845_disparity.npz
-NYUv2/test/kitchen_0040/rgb_0846_rgb_left.mp4,NYUv2/test/kitchen_0040/rgb_0846_disparity.npz
-NYUv2/test/kitchen_0042/rgb_0850_rgb_left.mp4,NYUv2/test/kitchen_0042/rgb_0850_disparity.npz
-NYUv2/test/kitchen_0042/rgb_0851_rgb_left.mp4,NYUv2/test/kitchen_0042/rgb_0851_disparity.npz
-NYUv2/test/kitchen_0042/rgb_0852_rgb_left.mp4,NYUv2/test/kitchen_0042/rgb_0852_disparity.npz
-NYUv2/test/kitchen_0044/rgb_0857_rgb_left.mp4,NYUv2/test/kitchen_0044/rgb_0857_disparity.npz
-NYUv2/test/kitchen_0044/rgb_0858_rgb_left.mp4,NYUv2/test/kitchen_0044/rgb_0858_disparity.npz
-NYUv2/test/kitchen_0044/rgb_0859_rgb_left.mp4,NYUv2/test/kitchen_0044/rgb_0859_disparity.npz
-NYUv2/test/kitchen_0044/rgb_0860_rgb_left.mp4,NYUv2/test/kitchen_0044/rgb_0860_disparity.npz
-NYUv2/test/kitchen_0044/rgb_0861_rgb_left.mp4,NYUv2/test/kitchen_0044/rgb_0861_disparity.npz
-NYUv2/test/kitchen_0044/rgb_0862_rgb_left.mp4,NYUv2/test/kitchen_0044/rgb_0862_disparity.npz
-NYUv2/test/kitchen_0046/rgb_0869_rgb_left.mp4,NYUv2/test/kitchen_0046/rgb_0869_disparity.npz
-NYUv2/test/kitchen_0046/rgb_0870_rgb_left.mp4,NYUv2/test/kitchen_0046/rgb_0870_disparity.npz
-NYUv2/test/kitchen_0046/rgb_0871_rgb_left.mp4,NYUv2/test/kitchen_0046/rgb_0871_disparity.npz
-NYUv2/test/kitchen_0054/rgb_0906_rgb_left.mp4,NYUv2/test/kitchen_0054/rgb_0906_disparity.npz
-NYUv2/test/kitchen_0054/rgb_0907_rgb_left.mp4,NYUv2/test/kitchen_0054/rgb_0907_disparity.npz
-NYUv2/test/kitchen_0054/rgb_0908_rgb_left.mp4,NYUv2/test/kitchen_0054/rgb_0908_disparity.npz
-NYUv2/test/bedroom_0027/rgb_0917_rgb_left.mp4,NYUv2/test/bedroom_0027/rgb_0917_disparity.npz
-NYUv2/test/bedroom_0027/rgb_0918_rgb_left.mp4,NYUv2/test/bedroom_0027/rgb_0918_disparity.npz
-NYUv2/test/bedroom_0027/rgb_0919_rgb_left.mp4,NYUv2/test/bedroom_0027/rgb_0919_disparity.npz
-NYUv2/test/bedroom_0030/rgb_0926_rgb_left.mp4,NYUv2/test/bedroom_0030/rgb_0926_disparity.npz
-NYUv2/test/bedroom_0030/rgb_0927_rgb_left.mp4,NYUv2/test/bedroom_0030/rgb_0927_disparity.npz
-NYUv2/test/bedroom_0030/rgb_0928_rgb_left.mp4,NYUv2/test/bedroom_0030/rgb_0928_disparity.npz
-NYUv2/test/bedroom_0032/rgb_0932_rgb_left.mp4,NYUv2/test/bedroom_0032/rgb_0932_disparity.npz
-NYUv2/test/bedroom_0032/rgb_0933_rgb_left.mp4,NYUv2/test/bedroom_0032/rgb_0933_disparity.npz
-NYUv2/test/bedroom_0032/rgb_0934_rgb_left.mp4,NYUv2/test/bedroom_0032/rgb_0934_disparity.npz
-NYUv2/test/bedroom_0032/rgb_0935_rgb_left.mp4,NYUv2/test/bedroom_0032/rgb_0935_disparity.npz
-NYUv2/test/bedroom_0037/rgb_0945_rgb_left.mp4,NYUv2/test/bedroom_0037/rgb_0945_disparity.npz
-NYUv2/test/bedroom_0037/rgb_0946_rgb_left.mp4,NYUv2/test/bedroom_0037/rgb_0946_disparity.npz
-NYUv2/test/bedroom_0037/rgb_0947_rgb_left.mp4,NYUv2/test/bedroom_0037/rgb_0947_disparity.npz
-NYUv2/test/bedroom_0043/rgb_0959_rgb_left.mp4,NYUv2/test/bedroom_0043/rgb_0959_disparity.npz
-NYUv2/test/bedroom_0043/rgb_0960_rgb_left.mp4,NYUv2/test/bedroom_0043/rgb_0960_disparity.npz
-NYUv2/test/bedroom_0044/rgb_0961_rgb_left.mp4,NYUv2/test/bedroom_0044/rgb_0961_disparity.npz
-NYUv2/test/bedroom_0044/rgb_0962_rgb_left.mp4,NYUv2/test/bedroom_0044/rgb_0962_disparity.npz
-NYUv2/test/bedroom_0046/rgb_0965_rgb_left.mp4,NYUv2/test/bedroom_0046/rgb_0965_disparity.npz
-NYUv2/test/bedroom_0046/rgb_0966_rgb_left.mp4,NYUv2/test/bedroom_0046/rgb_0966_disparity.npz
-NYUv2/test/bedroom_0046/rgb_0967_rgb_left.mp4,NYUv2/test/bedroom_0046/rgb_0967_disparity.npz
-NYUv2/test/bedroom_0048/rgb_0970_rgb_left.mp4,NYUv2/test/bedroom_0048/rgb_0970_disparity.npz
-NYUv2/test/bedroom_0048/rgb_0971_rgb_left.mp4,NYUv2/test/bedroom_0048/rgb_0971_disparity.npz
-NYUv2/test/bedroom_0048/rgb_0972_rgb_left.mp4,NYUv2/test/bedroom_0048/rgb_0972_disparity.npz
-NYUv2/test/bedroom_0048/rgb_0973_rgb_left.mp4,NYUv2/test/bedroom_0048/rgb_0973_disparity.npz
-NYUv2/test/bedroom_0048/rgb_0974_rgb_left.mp4,NYUv2/test/bedroom_0048/rgb_0974_disparity.npz
-NYUv2/test/bedroom_0049/rgb_0975_rgb_left.mp4,NYUv2/test/bedroom_0049/rgb_0975_disparity.npz
-NYUv2/test/bedroom_0049/rgb_0976_rgb_left.mp4,NYUv2/test/bedroom_0049/rgb_0976_disparity.npz
-NYUv2/test/bedroom_0049/rgb_0977_rgb_left.mp4,NYUv2/test/bedroom_0049/rgb_0977_disparity.npz
-NYUv2/test/bedroom_0054/rgb_0991_rgb_left.mp4,NYUv2/test/bedroom_0054/rgb_0991_disparity.npz
-NYUv2/test/bedroom_0054/rgb_0992_rgb_left.mp4,NYUv2/test/bedroom_0054/rgb_0992_disparity.npz
-NYUv2/test/bedroom_0054/rgb_0993_rgb_left.mp4,NYUv2/test/bedroom_0054/rgb_0993_disparity.npz
-NYUv2/test/bedroom_0055/rgb_0994_rgb_left.mp4,NYUv2/test/bedroom_0055/rgb_0994_disparity.npz
-NYUv2/test/bedroom_0055/rgb_0995_rgb_left.mp4,NYUv2/test/bedroom_0055/rgb_0995_disparity.npz
-NYUv2/test/bedroom_0058/rgb_1001_rgb_left.mp4,NYUv2/test/bedroom_0058/rgb_1001_disparity.npz
-NYUv2/test/bedroom_0058/rgb_1002_rgb_left.mp4,NYUv2/test/bedroom_0058/rgb_1002_disparity.npz
-NYUv2/test/bedroom_0058/rgb_1003_rgb_left.mp4,NYUv2/test/bedroom_0058/rgb_1003_disparity.npz
-NYUv2/test/bedroom_0058/rgb_1004_rgb_left.mp4,NYUv2/test/bedroom_0058/rgb_1004_disparity.npz
-NYUv2/test/bedroom_0061/rgb_1010_rgb_left.mp4,NYUv2/test/bedroom_0061/rgb_1010_disparity.npz
-NYUv2/test/bedroom_0061/rgb_1011_rgb_left.mp4,NYUv2/test/bedroom_0061/rgb_1011_disparity.npz
-NYUv2/test/bedroom_0061/rgb_1012_rgb_left.mp4,NYUv2/test/bedroom_0061/rgb_1012_disparity.npz
-NYUv2/test/bedroom_0064/rgb_1021_rgb_left.mp4,NYUv2/test/bedroom_0064/rgb_1021_disparity.npz
-NYUv2/test/bedroom_0064/rgb_1022_rgb_left.mp4,NYUv2/test/bedroom_0064/rgb_1022_disparity.npz
-NYUv2/test/bedroom_0064/rgb_1023_rgb_left.mp4,NYUv2/test/bedroom_0064/rgb_1023_disparity.npz
-NYUv2/test/bedroom_0068/rgb_1032_rgb_left.mp4,NYUv2/test/bedroom_0068/rgb_1032_disparity.npz
-NYUv2/test/bedroom_0068/rgb_1033_rgb_left.mp4,NYUv2/test/bedroom_0068/rgb_1033_disparity.npz
-NYUv2/test/bedroom_0068/rgb_1034_rgb_left.mp4,NYUv2/test/bedroom_0068/rgb_1034_disparity.npz
-NYUv2/test/bedroom_0070/rgb_1038_rgb_left.mp4,NYUv2/test/bedroom_0070/rgb_1038_disparity.npz
-NYUv2/test/bedroom_0070/rgb_1039_rgb_left.mp4,NYUv2/test/bedroom_0070/rgb_1039_disparity.npz
-NYUv2/test/bedroom_0073/rgb_1048_rgb_left.mp4,NYUv2/test/bedroom_0073/rgb_1048_disparity.npz
-NYUv2/test/bedroom_0073/rgb_1049_rgb_left.mp4,NYUv2/test/bedroom_0073/rgb_1049_disparity.npz
-NYUv2/test/bedroom_0075/rgb_1052_rgb_left.mp4,NYUv2/test/bedroom_0075/rgb_1052_disparity.npz
-NYUv2/test/bedroom_0075/rgb_1053_rgb_left.mp4,NYUv2/test/bedroom_0075/rgb_1053_disparity.npz
-NYUv2/test/bedroom_0077/rgb_1057_rgb_left.mp4,NYUv2/test/bedroom_0077/rgb_1057_disparity.npz
-NYUv2/test/bedroom_0077/rgb_1058_rgb_left.mp4,NYUv2/test/bedroom_0077/rgb_1058_disparity.npz
-NYUv2/test/bedroom_0083/rgb_1075_rgb_left.mp4,NYUv2/test/bedroom_0083/rgb_1075_disparity.npz
-NYUv2/test/bedroom_0083/rgb_1076_rgb_left.mp4,NYUv2/test/bedroom_0083/rgb_1076_disparity.npz
-NYUv2/test/bedroom_0084/rgb_1077_rgb_left.mp4,NYUv2/test/bedroom_0084/rgb_1077_disparity.npz
-NYUv2/test/bedroom_0084/rgb_1078_rgb_left.mp4,NYUv2/test/bedroom_0084/rgb_1078_disparity.npz
-NYUv2/test/bedroom_0084/rgb_1079_rgb_left.mp4,NYUv2/test/bedroom_0084/rgb_1079_disparity.npz
-NYUv2/test/bedroom_0084/rgb_1080_rgb_left.mp4,NYUv2/test/bedroom_0084/rgb_1080_disparity.npz
-NYUv2/test/bedroom_0085/rgb_1081_rgb_left.mp4,NYUv2/test/bedroom_0085/rgb_1081_disparity.npz
-NYUv2/test/bedroom_0085/rgb_1082_rgb_left.mp4,NYUv2/test/bedroom_0085/rgb_1082_disparity.npz
-NYUv2/test/bedroom_0085/rgb_1083_rgb_left.mp4,NYUv2/test/bedroom_0085/rgb_1083_disparity.npz
-NYUv2/test/bedroom_0085/rgb_1084_rgb_left.mp4,NYUv2/test/bedroom_0085/rgb_1084_disparity.npz
-NYUv2/test/bedroom_0087/rgb_1088_rgb_left.mp4,NYUv2/test/bedroom_0087/rgb_1088_disparity.npz
-NYUv2/test/bedroom_0087/rgb_1089_rgb_left.mp4,NYUv2/test/bedroom_0087/rgb_1089_disparity.npz
-NYUv2/test/bedroom_0087/rgb_1090_rgb_left.mp4,NYUv2/test/bedroom_0087/rgb_1090_disparity.npz
-NYUv2/test/bedroom_0088/rgb_1091_rgb_left.mp4,NYUv2/test/bedroom_0088/rgb_1091_disparity.npz
-NYUv2/test/bedroom_0088/rgb_1092_rgb_left.mp4,NYUv2/test/bedroom_0088/rgb_1092_disparity.npz
-NYUv2/test/bedroom_0088/rgb_1093_rgb_left.mp4,NYUv2/test/bedroom_0088/rgb_1093_disparity.npz
-NYUv2/test/bedroom_0089/rgb_1094_rgb_left.mp4,NYUv2/test/bedroom_0089/rgb_1094_disparity.npz
-NYUv2/test/bedroom_0089/rgb_1095_rgb_left.mp4,NYUv2/test/bedroom_0089/rgb_1095_disparity.npz
-NYUv2/test/bedroom_0089/rgb_1096_rgb_left.mp4,NYUv2/test/bedroom_0089/rgb_1096_disparity.npz
-NYUv2/test/bedroom_0091/rgb_1098_rgb_left.mp4,NYUv2/test/bedroom_0091/rgb_1098_disparity.npz
-NYUv2/test/bedroom_0091/rgb_1099_rgb_left.mp4,NYUv2/test/bedroom_0091/rgb_1099_disparity.npz
-NYUv2/test/bedroom_0092/rgb_1100_rgb_left.mp4,NYUv2/test/bedroom_0092/rgb_1100_disparity.npz
-NYUv2/test/bedroom_0092/rgb_1101_rgb_left.mp4,NYUv2/test/bedroom_0092/rgb_1101_disparity.npz
-NYUv2/test/bedroom_0092/rgb_1102_rgb_left.mp4,NYUv2/test/bedroom_0092/rgb_1102_disparity.npz
-NYUv2/test/bedroom_0093/rgb_1103_rgb_left.mp4,NYUv2/test/bedroom_0093/rgb_1103_disparity.npz
-NYUv2/test/bedroom_0093/rgb_1104_rgb_left.mp4,NYUv2/test/bedroom_0093/rgb_1104_disparity.npz
-NYUv2/test/bedroom_0095/rgb_1106_rgb_left.mp4,NYUv2/test/bedroom_0095/rgb_1106_disparity.npz
-NYUv2/test/bedroom_0095/rgb_1107_rgb_left.mp4,NYUv2/test/bedroom_0095/rgb_1107_disparity.npz
-NYUv2/test/bedroom_0095/rgb_1108_rgb_left.mp4,NYUv2/test/bedroom_0095/rgb_1108_disparity.npz
-NYUv2/test/bedroom_0095/rgb_1109_rgb_left.mp4,NYUv2/test/bedroom_0095/rgb_1109_disparity.npz
-NYUv2/test/bedroom_0099/rgb_1117_rgb_left.mp4,NYUv2/test/bedroom_0099/rgb_1117_disparity.npz
-NYUv2/test/bedroom_0099/rgb_1118_rgb_left.mp4,NYUv2/test/bedroom_0099/rgb_1118_disparity.npz
-NYUv2/test/bedroom_0099/rgb_1119_rgb_left.mp4,NYUv2/test/bedroom_0099/rgb_1119_disparity.npz
-NYUv2/test/bedroom_0101/rgb_1123_rgb_left.mp4,NYUv2/test/bedroom_0101/rgb_1123_disparity.npz
-NYUv2/test/bedroom_0101/rgb_1124_rgb_left.mp4,NYUv2/test/bedroom_0101/rgb_1124_disparity.npz
-NYUv2/test/bedroom_0101/rgb_1125_rgb_left.mp4,NYUv2/test/bedroom_0101/rgb_1125_disparity.npz
-NYUv2/test/bedroom_0101/rgb_1126_rgb_left.mp4,NYUv2/test/bedroom_0101/rgb_1126_disparity.npz
-NYUv2/test/bedroom_0102/rgb_1127_rgb_left.mp4,NYUv2/test/bedroom_0102/rgb_1127_disparity.npz
-NYUv2/test/bedroom_0102/rgb_1128_rgb_left.mp4,NYUv2/test/bedroom_0102/rgb_1128_disparity.npz
-NYUv2/test/bedroom_0103/rgb_1129_rgb_left.mp4,NYUv2/test/bedroom_0103/rgb_1129_disparity.npz
-NYUv2/test/bedroom_0103/rgb_1130_rgb_left.mp4,NYUv2/test/bedroom_0103/rgb_1130_disparity.npz
-NYUv2/test/bedroom_0103/rgb_1131_rgb_left.mp4,NYUv2/test/bedroom_0103/rgb_1131_disparity.npz
-NYUv2/test/bedroom_0105/rgb_1135_rgb_left.mp4,NYUv2/test/bedroom_0105/rgb_1135_disparity.npz
-NYUv2/test/bedroom_0105/rgb_1136_rgb_left.mp4,NYUv2/test/bedroom_0105/rgb_1136_disparity.npz
-NYUv2/test/bedroom_0108/rgb_1144_rgb_left.mp4,NYUv2/test/bedroom_0108/rgb_1144_disparity.npz
-NYUv2/test/bedroom_0108/rgb_1145_rgb_left.mp4,NYUv2/test/bedroom_0108/rgb_1145_disparity.npz
-NYUv2/test/bedroom_0108/rgb_1146_rgb_left.mp4,NYUv2/test/bedroom_0108/rgb_1146_disparity.npz
-NYUv2/test/bedroom_0109/rgb_1147_rgb_left.mp4,NYUv2/test/bedroom_0109/rgb_1147_disparity.npz
-NYUv2/test/bedroom_0109/rgb_1148_rgb_left.mp4,NYUv2/test/bedroom_0109/rgb_1148_disparity.npz
-NYUv2/test/bedroom_0109/rgb_1149_rgb_left.mp4,NYUv2/test/bedroom_0109/rgb_1149_disparity.npz
-NYUv2/test/bedroom_0110/rgb_1150_rgb_left.mp4,NYUv2/test/bedroom_0110/rgb_1150_disparity.npz
-NYUv2/test/bedroom_0110/rgb_1151_rgb_left.mp4,NYUv2/test/bedroom_0110/rgb_1151_disparity.npz
-NYUv2/test/bedroom_0110/rgb_1152_rgb_left.mp4,NYUv2/test/bedroom_0110/rgb_1152_disparity.npz
-NYUv2/test/bedroom_0111/rgb_1153_rgb_left.mp4,NYUv2/test/bedroom_0111/rgb_1153_disparity.npz
-NYUv2/test/bedroom_0111/rgb_1154_rgb_left.mp4,NYUv2/test/bedroom_0111/rgb_1154_disparity.npz
-NYUv2/test/bedroom_0111/rgb_1155_rgb_left.mp4,NYUv2/test/bedroom_0111/rgb_1155_disparity.npz
-NYUv2/test/bedroom_0112/rgb_1156_rgb_left.mp4,NYUv2/test/bedroom_0112/rgb_1156_disparity.npz
-NYUv2/test/bedroom_0112/rgb_1157_rgb_left.mp4,NYUv2/test/bedroom_0112/rgb_1157_disparity.npz
-NYUv2/test/bedroom_0112/rgb_1158_rgb_left.mp4,NYUv2/test/bedroom_0112/rgb_1158_disparity.npz
-NYUv2/test/bedroom_0114/rgb_1162_rgb_left.mp4,NYUv2/test/bedroom_0114/rgb_1162_disparity.npz
-NYUv2/test/bedroom_0114/rgb_1163_rgb_left.mp4,NYUv2/test/bedroom_0114/rgb_1163_disparity.npz
-NYUv2/test/bedroom_0114/rgb_1164_rgb_left.mp4,NYUv2/test/bedroom_0114/rgb_1164_disparity.npz
-NYUv2/test/bedroom_0115/rgb_1165_rgb_left.mp4,NYUv2/test/bedroom_0115/rgb_1165_disparity.npz
-NYUv2/test/bedroom_0115/rgb_1166_rgb_left.mp4,NYUv2/test/bedroom_0115/rgb_1166_disparity.npz
-NYUv2/test/bedroom_0115/rgb_1167_rgb_left.mp4,NYUv2/test/bedroom_0115/rgb_1167_disparity.npz
-NYUv2/test/bedroom_0117/rgb_1170_rgb_left.mp4,NYUv2/test/bedroom_0117/rgb_1170_disparity.npz
-NYUv2/test/bedroom_0117/rgb_1171_rgb_left.mp4,NYUv2/test/bedroom_0117/rgb_1171_disparity.npz
-NYUv2/test/bedroom_0119/rgb_1174_rgb_left.mp4,NYUv2/test/bedroom_0119/rgb_1174_disparity.npz
-NYUv2/test/bedroom_0119/rgb_1175_rgb_left.mp4,NYUv2/test/bedroom_0119/rgb_1175_disparity.npz
-NYUv2/test/bedroom_0119/rgb_1176_rgb_left.mp4,NYUv2/test/bedroom_0119/rgb_1176_disparity.npz
-NYUv2/test/bedroom_0121/rgb_1179_rgb_left.mp4,NYUv2/test/bedroom_0121/rgb_1179_disparity.npz
-NYUv2/test/bedroom_0121/rgb_1180_rgb_left.mp4,NYUv2/test/bedroom_0121/rgb_1180_disparity.npz
-NYUv2/test/bedroom_0122/rgb_1181_rgb_left.mp4,NYUv2/test/bedroom_0122/rgb_1181_disparity.npz
-NYUv2/test/bedroom_0122/rgb_1182_rgb_left.mp4,NYUv2/test/bedroom_0122/rgb_1182_disparity.npz
-NYUv2/test/bedroom_0123/rgb_1183_rgb_left.mp4,NYUv2/test/bedroom_0123/rgb_1183_disparity.npz
-NYUv2/test/bedroom_0123/rgb_1184_rgb_left.mp4,NYUv2/test/bedroom_0123/rgb_1184_disparity.npz
-NYUv2/test/bedroom_0127/rgb_1192_rgb_left.mp4,NYUv2/test/bedroom_0127/rgb_1192_disparity.npz
-NYUv2/test/bedroom_0127/rgb_1193_rgb_left.mp4,NYUv2/test/bedroom_0127/rgb_1193_disparity.npz
-NYUv2/test/bedroom_0127/rgb_1194_rgb_left.mp4,NYUv2/test/bedroom_0127/rgb_1194_disparity.npz
-NYUv2/test/bedroom_0128/rgb_1195_rgb_left.mp4,NYUv2/test/bedroom_0128/rgb_1195_disparity.npz
-NYUv2/test/bedroom_0128/rgb_1196_rgb_left.mp4,NYUv2/test/bedroom_0128/rgb_1196_disparity.npz
-NYUv2/test/living_room_0025/rgb_1201_rgb_left.mp4,NYUv2/test/living_room_0025/rgb_1201_disparity.npz
-NYUv2/test/living_room_0025/rgb_1202_rgb_left.mp4,NYUv2/test/living_room_0025/rgb_1202_disparity.npz
-NYUv2/test/living_room_0025/rgb_1203_rgb_left.mp4,NYUv2/test/living_room_0025/rgb_1203_disparity.npz
-NYUv2/test/living_room_0026/rgb_1204_rgb_left.mp4,NYUv2/test/living_room_0026/rgb_1204_disparity.npz
-NYUv2/test/living_room_0026/rgb_1205_rgb_left.mp4,NYUv2/test/living_room_0026/rgb_1205_disparity.npz
-NYUv2/test/living_room_0027/rgb_1206_rgb_left.mp4,NYUv2/test/living_room_0027/rgb_1206_disparity.npz
-NYUv2/test/living_room_0027/rgb_1207_rgb_left.mp4,NYUv2/test/living_room_0027/rgb_1207_disparity.npz
-NYUv2/test/living_room_0027/rgb_1208_rgb_left.mp4,NYUv2/test/living_room_0027/rgb_1208_disparity.npz
-NYUv2/test/living_room_0028/rgb_1209_rgb_left.mp4,NYUv2/test/living_room_0028/rgb_1209_disparity.npz
-NYUv2/test/living_room_0028/rgb_1210_rgb_left.mp4,NYUv2/test/living_room_0028/rgb_1210_disparity.npz
-NYUv2/test/living_room_0028/rgb_1211_rgb_left.mp4,NYUv2/test/living_room_0028/rgb_1211_disparity.npz
-NYUv2/test/living_room_0028/rgb_1212_rgb_left.mp4,NYUv2/test/living_room_0028/rgb_1212_disparity.npz
-NYUv2/test/living_room_0030/rgb_1216_rgb_left.mp4,NYUv2/test/living_room_0030/rgb_1216_disparity.npz
-NYUv2/test/living_room_0030/rgb_1217_rgb_left.mp4,NYUv2/test/living_room_0030/rgb_1217_disparity.npz
-NYUv2/test/living_room_0031/rgb_1218_rgb_left.mp4,NYUv2/test/living_room_0031/rgb_1218_disparity.npz
-NYUv2/test/living_room_0031/rgb_1219_rgb_left.mp4,NYUv2/test/living_room_0031/rgb_1219_disparity.npz
-NYUv2/test/living_room_0031/rgb_1220_rgb_left.mp4,NYUv2/test/living_room_0031/rgb_1220_disparity.npz
-NYUv2/test/living_room_0034/rgb_1226_rgb_left.mp4,NYUv2/test/living_room_0034/rgb_1226_disparity.npz
-NYUv2/test/living_room_0034/rgb_1227_rgb_left.mp4,NYUv2/test/living_room_0034/rgb_1227_disparity.npz
-NYUv2/test/living_room_0034/rgb_1228_rgb_left.mp4,NYUv2/test/living_room_0034/rgb_1228_disparity.npz
-NYUv2/test/living_room_0034/rgb_1229_rgb_left.mp4,NYUv2/test/living_room_0034/rgb_1229_disparity.npz
-NYUv2/test/living_room_0034/rgb_1230_rgb_left.mp4,NYUv2/test/living_room_0034/rgb_1230_disparity.npz
-NYUv2/test/living_room_0036/rgb_1233_rgb_left.mp4,NYUv2/test/living_room_0036/rgb_1233_disparity.npz
-NYUv2/test/living_room_0036/rgb_1234_rgb_left.mp4,NYUv2/test/living_room_0036/rgb_1234_disparity.npz
-NYUv2/test/living_room_0036/rgb_1235_rgb_left.mp4,NYUv2/test/living_room_0036/rgb_1235_disparity.npz
-NYUv2/test/living_room_0041/rgb_1247_rgb_left.mp4,NYUv2/test/living_room_0041/rgb_1247_disparity.npz
-NYUv2/test/living_room_0041/rgb_1248_rgb_left.mp4,NYUv2/test/living_room_0041/rgb_1248_disparity.npz
-NYUv2/test/living_room_0041/rgb_1249_rgb_left.mp4,NYUv2/test/living_room_0041/rgb_1249_disparity.npz
-NYUv2/test/living_room_0041/rgb_1250_rgb_left.mp4,NYUv2/test/living_room_0041/rgb_1250_disparity.npz
-NYUv2/test/living_room_0043/rgb_1254_rgb_left.mp4,NYUv2/test/living_room_0043/rgb_1254_disparity.npz
-NYUv2/test/living_room_0043/rgb_1255_rgb_left.mp4,NYUv2/test/living_room_0043/rgb_1255_disparity.npz
-NYUv2/test/living_room_0043/rgb_1256_rgb_left.mp4,NYUv2/test/living_room_0043/rgb_1256_disparity.npz
-NYUv2/test/living_room_0043/rgb_1257_rgb_left.mp4,NYUv2/test/living_room_0043/rgb_1257_disparity.npz
-NYUv2/test/living_room_0044/rgb_1258_rgb_left.mp4,NYUv2/test/living_room_0044/rgb_1258_disparity.npz
-NYUv2/test/living_room_0044/rgb_1259_rgb_left.mp4,NYUv2/test/living_room_0044/rgb_1259_disparity.npz
-NYUv2/test/living_room_0044/rgb_1260_rgb_left.mp4,NYUv2/test/living_room_0044/rgb_1260_disparity.npz
-NYUv2/test/living_room_0044/rgb_1261_rgb_left.mp4,NYUv2/test/living_room_0044/rgb_1261_disparity.npz
-NYUv2/test/living_room_0045/rgb_1262_rgb_left.mp4,NYUv2/test/living_room_0045/rgb_1262_disparity.npz
-NYUv2/test/living_room_0045/rgb_1263_rgb_left.mp4,NYUv2/test/living_room_0045/rgb_1263_disparity.npz
-NYUv2/test/living_room_0045/rgb_1264_rgb_left.mp4,NYUv2/test/living_room_0045/rgb_1264_disparity.npz
-NYUv2/test/living_room_0045/rgb_1265_rgb_left.mp4,NYUv2/test/living_room_0045/rgb_1265_disparity.npz
-NYUv2/test/living_room_0048/rgb_1275_rgb_left.mp4,NYUv2/test/living_room_0048/rgb_1275_disparity.npz
-NYUv2/test/living_room_0048/rgb_1276_rgb_left.mp4,NYUv2/test/living_room_0048/rgb_1276_disparity.npz
-NYUv2/test/living_room_0049/rgb_1277_rgb_left.mp4,NYUv2/test/living_room_0049/rgb_1277_disparity.npz
-NYUv2/test/living_room_0049/rgb_1278_rgb_left.mp4,NYUv2/test/living_room_0049/rgb_1278_disparity.npz
-NYUv2/test/living_room_0049/rgb_1279_rgb_left.mp4,NYUv2/test/living_room_0049/rgb_1279_disparity.npz
-NYUv2/test/living_room_0049/rgb_1280_rgb_left.mp4,NYUv2/test/living_room_0049/rgb_1280_disparity.npz
-NYUv2/test/living_room_0051/rgb_1285_rgb_left.mp4,NYUv2/test/living_room_0051/rgb_1285_disparity.npz
-NYUv2/test/living_room_0051/rgb_1286_rgb_left.mp4,NYUv2/test/living_room_0051/rgb_1286_disparity.npz
-NYUv2/test/living_room_0051/rgb_1287_rgb_left.mp4,NYUv2/test/living_room_0051/rgb_1287_disparity.npz
-NYUv2/test/living_room_0051/rgb_1288_rgb_left.mp4,NYUv2/test/living_room_0051/rgb_1288_disparity.npz
-NYUv2/test/living_room_0052/rgb_1289_rgb_left.mp4,NYUv2/test/living_room_0052/rgb_1289_disparity.npz
-NYUv2/test/living_room_0052/rgb_1290_rgb_left.mp4,NYUv2/test/living_room_0052/rgb_1290_disparity.npz
-NYUv2/test/living_room_0053/rgb_1291_rgb_left.mp4,NYUv2/test/living_room_0053/rgb_1291_disparity.npz
-NYUv2/test/living_room_0053/rgb_1292_rgb_left.mp4,NYUv2/test/living_room_0053/rgb_1292_disparity.npz
-NYUv2/test/living_room_0053/rgb_1293_rgb_left.mp4,NYUv2/test/living_room_0053/rgb_1293_disparity.npz
-NYUv2/test/living_room_0054/rgb_1294_rgb_left.mp4,NYUv2/test/living_room_0054/rgb_1294_disparity.npz
-NYUv2/test/living_room_0054/rgb_1295_rgb_left.mp4,NYUv2/test/living_room_0054/rgb_1295_disparity.npz
-NYUv2/test/living_room_0056/rgb_1297_rgb_left.mp4,NYUv2/test/living_room_0056/rgb_1297_disparity.npz
-NYUv2/test/living_room_0057/rgb_1298_rgb_left.mp4,NYUv2/test/living_room_0057/rgb_1298_disparity.npz
-NYUv2/test/living_room_0057/rgb_1299_rgb_left.mp4,NYUv2/test/living_room_0057/rgb_1299_disparity.npz
-NYUv2/test/living_room_0059/rgb_1302_rgb_left.mp4,NYUv2/test/living_room_0059/rgb_1302_disparity.npz
-NYUv2/test/living_room_0059/rgb_1303_rgb_left.mp4,NYUv2/test/living_room_0059/rgb_1303_disparity.npz
-NYUv2/test/living_room_0059/rgb_1304_rgb_left.mp4,NYUv2/test/living_room_0059/rgb_1304_disparity.npz
-NYUv2/test/living_room_0059/rgb_1305_rgb_left.mp4,NYUv2/test/living_room_0059/rgb_1305_disparity.npz
-NYUv2/test/living_room_0060/rgb_1306_rgb_left.mp4,NYUv2/test/living_room_0060/rgb_1306_disparity.npz
-NYUv2/test/living_room_0060/rgb_1307_rgb_left.mp4,NYUv2/test/living_room_0060/rgb_1307_disparity.npz
-NYUv2/test/living_room_0061/rgb_1308_rgb_left.mp4,NYUv2/test/living_room_0061/rgb_1308_disparity.npz
-NYUv2/test/living_room_0064/rgb_1314_rgb_left.mp4,NYUv2/test/living_room_0064/rgb_1314_disparity.npz
-NYUv2/test/living_room_0066/rgb_1315_rgb_left.mp4,NYUv2/test/living_room_0066/rgb_1315_disparity.npz
-NYUv2/test/living_room_0072/rgb_1329_rgb_left.mp4,NYUv2/test/living_room_0072/rgb_1329_disparity.npz
-NYUv2/test/living_room_0075/rgb_1330_rgb_left.mp4,NYUv2/test/living_room_0075/rgb_1330_disparity.npz
-NYUv2/test/living_room_0075/rgb_1331_rgb_left.mp4,NYUv2/test/living_room_0075/rgb_1331_disparity.npz
-NYUv2/test/living_room_0076/rgb_1332_rgb_left.mp4,NYUv2/test/living_room_0076/rgb_1332_disparity.npz
-NYUv2/test/living_room_0079/rgb_1335_rgb_left.mp4,NYUv2/test/living_room_0079/rgb_1335_disparity.npz
-NYUv2/test/living_room_0079/rgb_1336_rgb_left.mp4,NYUv2/test/living_room_0079/rgb_1336_disparity.npz
-NYUv2/test/living_room_0079/rgb_1337_rgb_left.mp4,NYUv2/test/living_room_0079/rgb_1337_disparity.npz
-NYUv2/test/living_room_0080/rgb_1338_rgb_left.mp4,NYUv2/test/living_room_0080/rgb_1338_disparity.npz
-NYUv2/test/living_room_0080/rgb_1339_rgb_left.mp4,NYUv2/test/living_room_0080/rgb_1339_disparity.npz
-NYUv2/test/living_room_0080/rgb_1340_rgb_left.mp4,NYUv2/test/living_room_0080/rgb_1340_disparity.npz
-NYUv2/test/dining_room_0003/rgb_1347_rgb_left.mp4,NYUv2/test/dining_room_0003/rgb_1347_disparity.npz
-NYUv2/test/dining_room_0003/rgb_1348_rgb_left.mp4,NYUv2/test/dining_room_0003/rgb_1348_disparity.npz
-NYUv2/test/dining_room_0003/rgb_1349_rgb_left.mp4,NYUv2/test/dining_room_0003/rgb_1349_disparity.npz
-NYUv2/test/dining_room_0005/rgb_1353_rgb_left.mp4,NYUv2/test/dining_room_0005/rgb_1353_disparity.npz
-NYUv2/test/dining_room_0005/rgb_1354_rgb_left.mp4,NYUv2/test/dining_room_0005/rgb_1354_disparity.npz
-NYUv2/test/dining_room_0006/rgb_1355_rgb_left.mp4,NYUv2/test/dining_room_0006/rgb_1355_disparity.npz
-NYUv2/test/dining_room_0006/rgb_1356_rgb_left.mp4,NYUv2/test/dining_room_0006/rgb_1356_disparity.npz
-NYUv2/test/dining_room_0009/rgb_1364_rgb_left.mp4,NYUv2/test/dining_room_0009/rgb_1364_disparity.npz
-NYUv2/test/dining_room_0009/rgb_1365_rgb_left.mp4,NYUv2/test/dining_room_0009/rgb_1365_disparity.npz
-NYUv2/test/dining_room_0011/rgb_1368_rgb_left.mp4,NYUv2/test/dining_room_0011/rgb_1368_disparity.npz
-NYUv2/test/dining_room_0011/rgb_1369_rgb_left.mp4,NYUv2/test/dining_room_0011/rgb_1369_disparity.npz
-NYUv2/test/dining_room_0017/rgb_1384_rgb_left.mp4,NYUv2/test/dining_room_0017/rgb_1384_disparity.npz
-NYUv2/test/dining_room_0017/rgb_1385_rgb_left.mp4,NYUv2/test/dining_room_0017/rgb_1385_disparity.npz
-NYUv2/test/dining_room_0017/rgb_1386_rgb_left.mp4,NYUv2/test/dining_room_0017/rgb_1386_disparity.npz
-NYUv2/test/dining_room_0018/rgb_1387_rgb_left.mp4,NYUv2/test/dining_room_0018/rgb_1387_disparity.npz
-NYUv2/test/dining_room_0018/rgb_1388_rgb_left.mp4,NYUv2/test/dining_room_0018/rgb_1388_disparity.npz
-NYUv2/test/dining_room_0018/rgb_1389_rgb_left.mp4,NYUv2/test/dining_room_0018/rgb_1389_disparity.npz
-NYUv2/test/dining_room_0018/rgb_1390_rgb_left.mp4,NYUv2/test/dining_room_0018/rgb_1390_disparity.npz
-NYUv2/test/dining_room_0018/rgb_1391_rgb_left.mp4,NYUv2/test/dining_room_0018/rgb_1391_disparity.npz
-NYUv2/test/dining_room_0020/rgb_1394_rgb_left.mp4,NYUv2/test/dining_room_0020/rgb_1394_disparity.npz
-NYUv2/test/dining_room_0020/rgb_1395_rgb_left.mp4,NYUv2/test/dining_room_0020/rgb_1395_disparity.npz
-NYUv2/test/dining_room_0020/rgb_1396_rgb_left.mp4,NYUv2/test/dining_room_0020/rgb_1396_disparity.npz
-NYUv2/test/dining_room_0021/rgb_1397_rgb_left.mp4,NYUv2/test/dining_room_0021/rgb_1397_disparity.npz
-NYUv2/test/dining_room_0021/rgb_1398_rgb_left.mp4,NYUv2/test/dining_room_0021/rgb_1398_disparity.npz
-NYUv2/test/dining_room_0021/rgb_1399_rgb_left.mp4,NYUv2/test/dining_room_0021/rgb_1399_disparity.npz
-NYUv2/test/dining_room_0022/rgb_1400_rgb_left.mp4,NYUv2/test/dining_room_0022/rgb_1400_disparity.npz
-NYUv2/test/dining_room_0022/rgb_1401_rgb_left.mp4,NYUv2/test/dining_room_0022/rgb_1401_disparity.npz
-NYUv2/test/dining_room_0025/rgb_1407_rgb_left.mp4,NYUv2/test/dining_room_0025/rgb_1407_disparity.npz
-NYUv2/test/dining_room_0025/rgb_1408_rgb_left.mp4,NYUv2/test/dining_room_0025/rgb_1408_disparity.npz
-NYUv2/test/dining_room_0025/rgb_1409_rgb_left.mp4,NYUv2/test/dining_room_0025/rgb_1409_disparity.npz
-NYUv2/test/dining_room_0025/rgb_1410_rgb_left.mp4,NYUv2/test/dining_room_0025/rgb_1410_disparity.npz
-NYUv2/test/dining_room_0025/rgb_1411_rgb_left.mp4,NYUv2/test/dining_room_0025/rgb_1411_disparity.npz
-NYUv2/test/dining_room_0026/rgb_1412_rgb_left.mp4,NYUv2/test/dining_room_0026/rgb_1412_disparity.npz
-NYUv2/test/dining_room_0026/rgb_1413_rgb_left.mp4,NYUv2/test/dining_room_0026/rgb_1413_disparity.npz
-NYUv2/test/dining_room_0026/rgb_1414_rgb_left.mp4,NYUv2/test/dining_room_0026/rgb_1414_disparity.npz
-NYUv2/test/dining_room_0030/rgb_1421_rgb_left.mp4,NYUv2/test/dining_room_0030/rgb_1421_disparity.npz
-NYUv2/test/dining_room_0030/rgb_1422_rgb_left.mp4,NYUv2/test/dining_room_0030/rgb_1422_disparity.npz
-NYUv2/test/dining_room_0030/rgb_1423_rgb_left.mp4,NYUv2/test/dining_room_0030/rgb_1423_disparity.npz
-NYUv2/test/dining_room_0030/rgb_1424_rgb_left.mp4,NYUv2/test/dining_room_0030/rgb_1424_disparity.npz
-NYUv2/test/dining_room_0032/rgb_1430_rgb_left.mp4,NYUv2/test/dining_room_0032/rgb_1430_disparity.npz
-NYUv2/test/dining_room_0032/rgb_1431_rgb_left.mp4,NYUv2/test/dining_room_0032/rgb_1431_disparity.npz
-NYUv2/test/dining_room_0032/rgb_1432_rgb_left.mp4,NYUv2/test/dining_room_0032/rgb_1432_disparity.npz
-NYUv2/test/dining_room_0032/rgb_1433_rgb_left.mp4,NYUv2/test/dining_room_0032/rgb_1433_disparity.npz
-NYUv2/test/dining_room_0035/rgb_1441_rgb_left.mp4,NYUv2/test/dining_room_0035/rgb_1441_disparity.npz
-NYUv2/test/dining_room_0035/rgb_1442_rgb_left.mp4,NYUv2/test/dining_room_0035/rgb_1442_disparity.npz
-NYUv2/test/dining_room_0035/rgb_1443_rgb_left.mp4,NYUv2/test/dining_room_0035/rgb_1443_disparity.npz
-NYUv2/test/dining_room_0036/rgb_1444_rgb_left.mp4,NYUv2/test/dining_room_0036/rgb_1444_disparity.npz
-NYUv2/test/dining_room_0036/rgb_1445_rgb_left.mp4,NYUv2/test/dining_room_0036/rgb_1445_disparity.npz
-NYUv2/test/dining_room_0036/rgb_1446_rgb_left.mp4,NYUv2/test/dining_room_0036/rgb_1446_disparity.npz
-NYUv2/test/dining_room_0036/rgb_1447_rgb_left.mp4,NYUv2/test/dining_room_0036/rgb_1447_disparity.npz
-NYUv2/test/dining_room_0036/rgb_1448_rgb_left.mp4,NYUv2/test/dining_room_0036/rgb_1448_disparity.npz
-NYUv2/test/dining_room_0036/rgb_1449_rgb_left.mp4,NYUv2/test/dining_room_0036/rgb_1449_disparity.npz

benchmark/csv/meta_scannet_test.csv

@@ -1,101 +0,0 @@
-filepath_left,filepath_disparity
-scannet/scene0707_00_rgb_left.mp4,scannet/scene0707_00_disparity.npz
-scannet/scene0708_00_rgb_left.mp4,scannet/scene0708_00_disparity.npz
-scannet/scene0709_00_rgb_left.mp4,scannet/scene0709_00_disparity.npz
-scannet/scene0710_00_rgb_left.mp4,scannet/scene0710_00_disparity.npz
-scannet/scene0711_00_rgb_left.mp4,scannet/scene0711_00_disparity.npz
-scannet/scene0712_00_rgb_left.mp4,scannet/scene0712_00_disparity.npz
-scannet/scene0713_00_rgb_left.mp4,scannet/scene0713_00_disparity.npz
-scannet/scene0714_00_rgb_left.mp4,scannet/scene0714_00_disparity.npz
-scannet/scene0715_00_rgb_left.mp4,scannet/scene0715_00_disparity.npz
-scannet/scene0716_00_rgb_left.mp4,scannet/scene0716_00_disparity.npz
-scannet/scene0717_00_rgb_left.mp4,scannet/scene0717_00_disparity.npz
-scannet/scene0718_00_rgb_left.mp4,scannet/scene0718_00_disparity.npz
-scannet/scene0719_00_rgb_left.mp4,scannet/scene0719_00_disparity.npz
-scannet/scene0720_00_rgb_left.mp4,scannet/scene0720_00_disparity.npz
-scannet/scene0721_00_rgb_left.mp4,scannet/scene0721_00_disparity.npz
-scannet/scene0722_00_rgb_left.mp4,scannet/scene0722_00_disparity.npz
-scannet/scene0723_00_rgb_left.mp4,scannet/scene0723_00_disparity.npz
-scannet/scene0724_00_rgb_left.mp4,scannet/scene0724_00_disparity.npz
-scannet/scene0725_00_rgb_left.mp4,scannet/scene0725_00_disparity.npz
-scannet/scene0726_00_rgb_left.mp4,scannet/scene0726_00_disparity.npz
-scannet/scene0727_00_rgb_left.mp4,scannet/scene0727_00_disparity.npz
-scannet/scene0728_00_rgb_left.mp4,scannet/scene0728_00_disparity.npz
-scannet/scene0729_00_rgb_left.mp4,scannet/scene0729_00_disparity.npz
-scannet/scene0730_00_rgb_left.mp4,scannet/scene0730_00_disparity.npz
-scannet/scene0731_00_rgb_left.mp4,scannet/scene0731_00_disparity.npz
-scannet/scene0732_00_rgb_left.mp4,scannet/scene0732_00_disparity.npz
-scannet/scene0733_00_rgb_left.mp4,scannet/scene0733_00_disparity.npz
-scannet/scene0734_00_rgb_left.mp4,scannet/scene0734_00_disparity.npz
-scannet/scene0735_00_rgb_left.mp4,scannet/scene0735_00_disparity.npz
-scannet/scene0736_00_rgb_left.mp4,scannet/scene0736_00_disparity.npz
-scannet/scene0737_00_rgb_left.mp4,scannet/scene0737_00_disparity.npz
-scannet/scene0738_00_rgb_left.mp4,scannet/scene0738_00_disparity.npz
-scannet/scene0739_00_rgb_left.mp4,scannet/scene0739_00_disparity.npz
-scannet/scene0740_00_rgb_left.mp4,scannet/scene0740_00_disparity.npz
-scannet/scene0741_00_rgb_left.mp4,scannet/scene0741_00_disparity.npz
-scannet/scene0742_00_rgb_left.mp4,scannet/scene0742_00_disparity.npz
-scannet/scene0743_00_rgb_left.mp4,scannet/scene0743_00_disparity.npz
-scannet/scene0744_00_rgb_left.mp4,scannet/scene0744_00_disparity.npz
-scannet/scene0745_00_rgb_left.mp4,scannet/scene0745_00_disparity.npz
-scannet/scene0746_00_rgb_left.mp4,scannet/scene0746_00_disparity.npz
-scannet/scene0747_00_rgb_left.mp4,scannet/scene0747_00_disparity.npz
-scannet/scene0748_00_rgb_left.mp4,scannet/scene0748_00_disparity.npz
-scannet/scene0749_00_rgb_left.mp4,scannet/scene0749_00_disparity.npz
-scannet/scene0750_00_rgb_left.mp4,scannet/scene0750_00_disparity.npz
-scannet/scene0751_00_rgb_left.mp4,scannet/scene0751_00_disparity.npz
-scannet/scene0752_00_rgb_left.mp4,scannet/scene0752_00_disparity.npz
-scannet/scene0753_00_rgb_left.mp4,scannet/scene0753_00_disparity.npz
-scannet/scene0754_00_rgb_left.mp4,scannet/scene0754_00_disparity.npz
-scannet/scene0755_00_rgb_left.mp4,scannet/scene0755_00_disparity.npz
-scannet/scene0756_00_rgb_left.mp4,scannet/scene0756_00_disparity.npz
-scannet/scene0757_00_rgb_left.mp4,scannet/scene0757_00_disparity.npz
-scannet/scene0758_00_rgb_left.mp4,scannet/scene0758_00_disparity.npz
-scannet/scene0759_00_rgb_left.mp4,scannet/scene0759_00_disparity.npz
-scannet/scene0760_00_rgb_left.mp4,scannet/scene0760_00_disparity.npz
-scannet/scene0761_00_rgb_left.mp4,scannet/scene0761_00_disparity.npz
-scannet/scene0762_00_rgb_left.mp4,scannet/scene0762_00_disparity.npz
-scannet/scene0763_00_rgb_left.mp4,scannet/scene0763_00_disparity.npz
-scannet/scene0764_00_rgb_left.mp4,scannet/scene0764_00_disparity.npz
-scannet/scene0765_00_rgb_left.mp4,scannet/scene0765_00_disparity.npz
-scannet/scene0766_00_rgb_left.mp4,scannet/scene0766_00_disparity.npz
-scannet/scene0767_00_rgb_left.mp4,scannet/scene0767_00_disparity.npz
-scannet/scene0768_00_rgb_left.mp4,scannet/scene0768_00_disparity.npz
-scannet/scene0769_00_rgb_left.mp4,scannet/scene0769_00_disparity.npz
-scannet/scene0770_00_rgb_left.mp4,scannet/scene0770_00_disparity.npz
-scannet/scene0771_00_rgb_left.mp4,scannet/scene0771_00_disparity.npz
-scannet/scene0772_00_rgb_left.mp4,scannet/scene0772_00_disparity.npz
-scannet/scene0773_00_rgb_left.mp4,scannet/scene0773_00_disparity.npz
-scannet/scene0774_00_rgb_left.mp4,scannet/scene0774_00_disparity.npz
-scannet/scene0775_00_rgb_left.mp4,scannet/scene0775_00_disparity.npz
-scannet/scene0776_00_rgb_left.mp4,scannet/scene0776_00_disparity.npz
-scannet/scene0777_00_rgb_left.mp4,scannet/scene0777_00_disparity.npz
-scannet/scene0778_00_rgb_left.mp4,scannet/scene0778_00_disparity.npz
-scannet/scene0779_00_rgb_left.mp4,scannet/scene0779_00_disparity.npz
-scannet/scene0780_00_rgb_left.mp4,scannet/scene0780_00_disparity.npz
-scannet/scene0781_00_rgb_left.mp4,scannet/scene0781_00_disparity.npz
-scannet/scene0782_00_rgb_left.mp4,scannet/scene0782_00_disparity.npz
-scannet/scene0783_00_rgb_left.mp4,scannet/scene0783_00_disparity.npz
-scannet/scene0784_00_rgb_left.mp4,scannet/scene0784_00_disparity.npz
-scannet/scene0785_00_rgb_left.mp4,scannet/scene0785_00_disparity.npz
-scannet/scene0786_00_rgb_left.mp4,scannet/scene0786_00_disparity.npz
-scannet/scene0787_00_rgb_left.mp4,scannet/scene0787_00_disparity.npz
-scannet/scene0788_00_rgb_left.mp4,scannet/scene0788_00_disparity.npz
-scannet/scene0789_00_rgb_left.mp4,scannet/scene0789_00_disparity.npz
-scannet/scene0790_00_rgb_left.mp4,scannet/scene0790_00_disparity.npz
-scannet/scene0791_00_rgb_left.mp4,scannet/scene0791_00_disparity.npz
-scannet/scene0792_00_rgb_left.mp4,scannet/scene0792_00_disparity.npz
-scannet/scene0793_00_rgb_left.mp4,scannet/scene0793_00_disparity.npz
-scannet/scene0794_00_rgb_left.mp4,scannet/scene0794_00_disparity.npz
-scannet/scene0795_00_rgb_left.mp4,scannet/scene0795_00_disparity.npz
-scannet/scene0796_00_rgb_left.mp4,scannet/scene0796_00_disparity.npz
-scannet/scene0797_00_rgb_left.mp4,scannet/scene0797_00_disparity.npz
-scannet/scene0798_00_rgb_left.mp4,scannet/scene0798_00_disparity.npz
-scannet/scene0799_00_rgb_left.mp4,scannet/scene0799_00_disparity.npz
-scannet/scene0800_00_rgb_left.mp4,scannet/scene0800_00_disparity.npz
-scannet/scene0801_00_rgb_left.mp4,scannet/scene0801_00_disparity.npz
-scannet/scene0802_00_rgb_left.mp4,scannet/scene0802_00_disparity.npz
-scannet/scene0803_00_rgb_left.mp4,scannet/scene0803_00_disparity.npz
-scannet/scene0804_00_rgb_left.mp4,scannet/scene0804_00_disparity.npz
-scannet/scene0805_00_rgb_left.mp4,scannet/scene0805_00_disparity.npz
-scannet/scene0806_00_rgb_left.mp4,scannet/scene0806_00_disparity.npz

benchmark/csv/meta_sintel.csv

@@ -1,24 +0,0 @@
-filepath_left,filepath_disparity
-sintel/ambush_5_rgb_left.mp4,sintel/ambush_5_disparity.npz
-sintel/bamboo_2_rgb_left.mp4,sintel/bamboo_2_disparity.npz
-sintel/mountain_1_rgb_left.mp4,sintel/mountain_1_disparity.npz
-sintel/bamboo_1_rgb_left.mp4,sintel/bamboo_1_disparity.npz
-sintel/shaman_2_rgb_left.mp4,sintel/shaman_2_disparity.npz
-sintel/ambush_6_rgb_left.mp4,sintel/ambush_6_disparity.npz
-sintel/bandage_1_rgb_left.mp4,sintel/bandage_1_disparity.npz
-sintel/alley_1_rgb_left.mp4,sintel/alley_1_disparity.npz
-sintel/temple_3_rgb_left.mp4,sintel/temple_3_disparity.npz
-sintel/shaman_3_rgb_left.mp4,sintel/shaman_3_disparity.npz
-sintel/ambush_2_rgb_left.mp4,sintel/ambush_2_disparity.npz
-sintel/cave_4_rgb_left.mp4,sintel/cave_4_disparity.npz
-sintel/cave_2_rgb_left.mp4,sintel/cave_2_disparity.npz
-sintel/alley_2_rgb_left.mp4,sintel/alley_2_disparity.npz
-sintel/market_5_rgb_left.mp4,sintel/market_5_disparity.npz
-sintel/sleeping_2_rgb_left.mp4,sintel/sleeping_2_disparity.npz
-sintel/ambush_4_rgb_left.mp4,sintel/ambush_4_disparity.npz
-sintel/sleeping_1_rgb_left.mp4,sintel/sleeping_1_disparity.npz
-sintel/market_6_rgb_left.mp4,sintel/market_6_disparity.npz
-sintel/market_2_rgb_left.mp4,sintel/market_2_disparity.npz
-sintel/bandage_2_rgb_left.mp4,sintel/bandage_2_disparity.npz
-sintel/ambush_7_rgb_left.mp4,sintel/ambush_7_disparity.npz
-sintel/temple_2_rgb_left.mp4,sintel/temple_2_disparity.npz

benchmark/dataset_extract/dataset_extract_bonn.py

@@ -1,155 +0,0 @@
-import os
-import numpy as np
-import os.path as osp
-from PIL import Image
-from tqdm import tqdm
-import imageio
-import csv
-
-
-def depth_read(filename):
-    # loads depth map D from png file
-    # and returns it as a numpy array
-
-    depth_png = np.asarray(Image.open(filename))
-    # make sure we have a proper 16bit depth map here.. not 8bit!
-    assert np.max(depth_png) > 255
-
-    depth = depth_png.astype(np.float64) / 5000.0
-    depth[depth_png == 0] = -1.0
-    return depth
-
-
-def extract_bonn(
-    root,
-    depth_root,
-    sample_len=-1,
-    csv_save_path="",
-    datatset_name="",
-    saved_rgb_dir="",
-    saved_disp_dir="",
-    start_frame=0,
-    end_frame=110,
-):
-    scenes_names = os.listdir(depth_root)
-    all_samples = []
-    for i, seq_name in enumerate(tqdm(scenes_names)):
-        # load all images
-        all_img_names = os.listdir(osp.join(depth_root, seq_name, "rgb"))
-        all_img_names = [x for x in all_img_names if x.endswith(".png")]
-        print(f"sequence frame number: {len(all_img_names)}")
-
-        # for not zero padding image name
-        all_img_names.sort()
-        all_img_names = sorted(all_img_names, key=lambda x: int(x.split(".")[0][-4:]))
-        all_img_names = all_img_names[start_frame:end_frame]
-
-        all_depth_names = os.listdir(osp.join(depth_root, seq_name, "depth"))
-        all_depth_names = [x for x in all_depth_names if x.endswith(".png")]
-        print(f"sequence depth number: {len(all_depth_names)}")
-
-        # for not zero padding image name
-        all_depth_names.sort()
-        all_depth_names = sorted(
-            all_depth_names, key=lambda x: int(x.split(".")[0][-4:])
-        )
-        all_depth_names = all_depth_names[start_frame:end_frame]
-
-        seq_len = len(all_img_names)
-        step = sample_len if sample_len > 0 else seq_len
-
-        for ref_idx in range(0, seq_len, step):
-            print(f"Progress: {seq_name}, {ref_idx // step + 1} / {seq_len//step}")
-
-            video_imgs = []
-            video_depths = []
-
-            if (ref_idx + step) <= seq_len:
-                ref_e = ref_idx + step
-            else:
-                continue
-
-            # for idx in range(ref_idx, ref_idx + step):
-            for idx in range(ref_idx, ref_e):
-                im_path = osp.join(root, seq_name, "rgb", all_img_names[idx])
-                depth_path = osp.join(
-                    depth_root, seq_name, "depth", all_depth_names[idx]
-                )
-
-                depth = depth_read(depth_path)
-                disp = depth
-
-                video_depths.append(disp)
-                video_imgs.append(np.array(Image.open(im_path)))
-
-            disp_video = np.array(video_depths)[:, None]  # [:, 0:1, :, :, 0]
-            img_video = np.array(video_imgs)[..., 0:3]  # [:, 0, :, :, 0:3]
-
-            print(disp_video.max(), disp_video.min())
-
-            def even_or_odd(num):
-                if num % 2 == 0:
-                    return num
-                else:
-                    return num - 1
-
-            # print(disp_video.shape)
-            # print(img_video.shape)
-            height = disp_video.shape[-2]
-            width = disp_video.shape[-1]
-            height = even_or_odd(height)
-            width = even_or_odd(width)
-            disp_video = disp_video[:, :, 0:height, 0:width]
-            img_video = img_video[:, 0:height, 0:width]
-
-            data_root = saved_rgb_dir + datatset_name
-            disp_root = saved_disp_dir + datatset_name
-            os.makedirs(data_root, exist_ok=True)
-            os.makedirs(disp_root, exist_ok=True)
-
-            img_video_dir = data_root
-            disp_video_dir = disp_root
-
-            img_video_path = os.path.join(img_video_dir, f"{seq_name}_rgb_left.mp4")
-            disp_video_path = os.path.join(disp_video_dir, f"{seq_name}_disparity.npz")
-
-            imageio.mimsave(
-                img_video_path, img_video, fps=15, quality=9, macro_block_size=1
-            )
-            np.savez(disp_video_path, disparity=disp_video)
-
-            sample = {}
-            sample["filepath_left"] = os.path.join(
-                f"{datatset_name}/{seq_name}_rgb_left.mp4"
-            )  # img_video_path
-            sample["filepath_disparity"] = os.path.join(
-                f"{datatset_name}/{seq_name}_disparity.npz"
-            )  # disp_video_path
-
-            all_samples.append(sample)
-
-    # save csv file
-
-    filename_ = csv_save_path
-    os.makedirs(os.path.dirname(filename_), exist_ok=True)
-    fields = ["filepath_left", "filepath_disparity"]
-    with open(filename_, "w") as csvfile:
-        writer = csv.DictWriter(csvfile, fieldnames=fields)
-        writer.writeheader()
-        writer.writerows(all_samples)
-
-    print(f"{filename_} has been saved.")
-
-
-if __name__ == "__main__":
-    extract_bonn(
-        root="path/to/Bonn-RGBD",
-        depth_root="path/to/Bonn-RGBD",
-        saved_rgb_dir="./benchmark/datasets/",
-        saved_disp_dir="./benchmark/datasets/",
-        csv_save_path=f"./benchmark/datasets/bonn.csv",
-        sample_len=-1,
-        datatset_name="bonn",
-        start_frame=30,
-        end_frame=140,
-    )

benchmark/dataset_extract/dataset_extract_kitti.py

@@ -1,140 +0,0 @@
-import os
-import numpy as np
-import os.path as osp
-from PIL import Image
-from tqdm import tqdm
-import csv
-import imageio
-
-
-def depth_read(filename):
-    # loads depth map D from png file
-    # and returns it as a numpy array,
-
-    depth_png = np.array(Image.open(filename), dtype=int)
-    # make sure we have a proper 16bit depth map here.. not 8bit!
-    assert np.max(depth_png) > 255
-
-    depth = depth_png.astype(np.float64) / 256.0
-    depth[depth_png == 0] = -1.0
-    return depth
-
-
-def extract_kitti(
-    root,
-    depth_root,
-    sample_len=-1,
-    csv_save_path="",
-    datatset_name="",
-    saved_rgb_dir="",
-    saved_disp_dir="",
-    start_frame=0,
-    end_frame=110,
-):
-    scenes_names = os.listdir(depth_root)
-    all_samples = []
-    for i, seq_name in enumerate(tqdm(scenes_names)):
-        all_img_names = os.listdir(
-            osp.join(depth_root, seq_name, "proj_depth/groundtruth/image_02")
-        )
-        all_img_names = [x for x in all_img_names if x.endswith(".png")]
-        print(f"sequence frame number: {len(all_img_names)}")
-
-        all_img_names.sort()
-        all_img_names = sorted(all_img_names, key=lambda x: int(x.split(".")[0][-4:]))
-        all_img_names = all_img_names[start_frame:end_frame]
-
-        seq_len = len(all_img_names)
-        step = sample_len if sample_len > 0 else seq_len
-
-        for ref_idx in range(0, seq_len, step):
-            print(f"Progress: {seq_name}, {ref_idx // step + 1} / {seq_len//step}")
-
-            video_imgs = []
-            video_depths = []
-
-            if (ref_idx + step) <= seq_len:
-                ref_e = ref_idx + step
-            else:
-                continue
-
-            for idx in range(ref_idx, ref_e):
-                im_path = osp.join(
-                    root, seq_name[0:10], seq_name, "image_02/data", all_img_names[idx]
-                )
-                depth_path = osp.join(
-                    depth_root,
-                    seq_name,
-                    "proj_depth/groundtruth/image_02",
-                    all_img_names[idx],
-                )
-
-                depth = depth_read(depth_path)
-                disp = depth
-
-                video_depths.append(disp)
-                video_imgs.append(np.array(Image.open(im_path)))
-
-            disp_video = np.array(video_depths)[:, None]
-            img_video = np.array(video_imgs)[..., 0:3]
-
-            def even_or_odd(num):
-                if num % 2 == 0:
-                    return num
-                else:
-                    return num - 1
-
-            height = disp_video.shape[-2]
-            width = disp_video.shape[-1]
-            height = even_or_odd(height)
-            width = even_or_odd(width)
-            disp_video = disp_video[:, :, 0:height, 0:width]
-            img_video = img_video[:, 0:height, 0:width]
-
-            data_root = saved_rgb_dir + datatset_name
-            disp_root = saved_disp_dir + datatset_name
-            os.makedirs(data_root, exist_ok=True)
-            os.makedirs(disp_root, exist_ok=True)
-
-            img_video_dir = data_root
-            disp_video_dir = disp_root
-
-            img_video_path = os.path.join(img_video_dir, f"{seq_name}_rgb_left.mp4")
-            disp_video_path = os.path.join(disp_video_dir, f"{seq_name}_disparity.npz")
-
-            imageio.mimsave(
-                img_video_path, img_video, fps=15, quality=10, macro_block_size=1
-            )
-            np.savez(disp_video_path, disparity=disp_video)
-
-            sample = {}
-            sample["filepath_left"] = os.path.join(f"KITTI/{seq_name}_rgb_left.mp4")
-            sample["filepath_disparity"] = os.path.join(
-                f"KITTI/{seq_name}_disparity.npz"
-            )
-
-            all_samples.append(sample)
-
-    filename_ = csv_save_path
-    os.makedirs(os.path.dirname(filename_), exist_ok=True)
-    fields = ["filepath_left", "filepath_disparity"]
-    with open(filename_, "w") as csvfile:
-        writer = csv.DictWriter(csvfile, fieldnames=fields)
-        writer.writeheader()
-        writer.writerows(all_samples)
-
-    print(f"{filename_} has been saved.")
-
-
-if __name__ == "__main__":
-    extract_kitti(
-        root="path/to/KITTI/raw_data",
-        depth_root="path/to/KITTI/data_depth_annotated/val",
-        saved_rgb_dir="./benchmark/datasets/",
-        saved_disp_dir="./benchmark/datasets/",
-        csv_save_path=f"./benchmark/datasets/KITTI.csv",
-        sample_len=-1,
-        datatset_name="KITTI",
-        start_frame=0,
-        end_frame=110,
-    )

benchmark/dataset_extract/dataset_extract_nyu.py

@@ -1,106 +0,0 @@
-import os
-import numpy as np
-import os.path as osp
-from PIL import Image
-from tqdm import tqdm
-import csv
-import imageio
-
-
-def _read_image(img_rel_path) -> np.ndarray:
-    image_to_read = img_rel_path
-    image = Image.open(image_to_read)
-    image = np.asarray(image)
-    return image
-
-
-def depth_read(filename):
-    depth_in = _read_image(filename)
-    depth_decoded = depth_in / 1000.0
-    return depth_decoded
-
-
-def extract_nyu(
-    root,
-    depth_root,
-    csv_save_path="",
-    datatset_name="",
-    filename_ls_path="",
-    saved_rgb_dir="",
-    saved_disp_dir="",
-):
-    with open(filename_ls_path, "r") as f:
-        filenames = [s.split() for s in f.readlines()]
-
-    all_samples = []
-    for i, pair_names in enumerate(tqdm(filenames)):
-        img_name = pair_names[0]
-        filled_depth_name = pair_names[2]
-
-        im_path = osp.join(root, img_name)
-        depth_path = osp.join(depth_root, filled_depth_name)
-
-        depth = depth_read(depth_path)
-        disp = depth
-
-        video_depths = [disp]
-        video_imgs = [np.array(Image.open(im_path))]
-
-        disp_video = np.array(video_depths)[:, None]
-        img_video = np.array(video_imgs)[..., 0:3]
-
-        disp_video = disp_video[:, :, 45:471, 41:601]
-        img_video = img_video[:, 45:471, 41:601, :]
-
-        data_root = saved_rgb_dir + datatset_name
-        disp_root = saved_disp_dir + datatset_name
-        os.makedirs(data_root, exist_ok=True)
-        os.makedirs(disp_root, exist_ok=True)
-
-        img_video_dir = data_root
-        disp_video_dir = disp_root
-
-        img_video_path = os.path.join(img_video_dir, f"{img_name[:-4]}_rgb_left.mp4")
-        disp_video_path = os.path.join(disp_video_dir, f"{img_name[:-4]}_disparity.npz")
-
-        dir_name = os.path.dirname(img_video_path)
-        os.makedirs(dir_name, exist_ok=True)
-        dir_name = os.path.dirname(disp_video_path)
-        os.makedirs(dir_name, exist_ok=True)
-
-        imageio.mimsave(
-            img_video_path, img_video, fps=15, quality=10, macro_block_size=1
-        )
-        np.savez(disp_video_path, disparity=disp_video)
-
-        sample = {}
-        sample["filepath_left"] = os.path.join(
-            f"{datatset_name}/{img_name[:-4]}_rgb_left.mp4"
-        )
-        sample["filepath_disparity"] = os.path.join(
-            f"{datatset_name}/{img_name[:-4]}_disparity.npz"
-        )
-
-        all_samples.append(sample)
-
-    filename_ = csv_save_path
-    os.makedirs(os.path.dirname(filename_), exist_ok=True)
-    fields = ["filepath_left", "filepath_disparity"]
-    with open(filename_, "w") as csvfile:
-        writer = csv.DictWriter(csvfile, fieldnames=fields)
-        writer.writeheader()
-        writer.writerows(all_samples)
-
-    print(f"{filename_} has been saved.")
-
-
-if __name__ == "__main__":
-    extract_nyu(
-        root="path/to/NYUv2/",
-        depth_root="path/to/NYUv2/",
-        filename_ls_path="path/to/NYUv2/filename_list_test.txt",
-        saved_rgb_dir="./benchmark/datasets/",
-        saved_disp_dir="./benchmark/datasets/",
-        csv_save_path=f"./benchmark/datasets/NYUv2.csv",
-        datatset_name="NYUv2",
-    )

benchmark/dataset_extract/dataset_extract_scannet.py

@@ -1,124 +0,0 @@
-import os
-import numpy as np
-import os.path as osp
-from PIL import Image
-from tqdm import tqdm
-import csv
-import imageio
-
-
-def _read_image(img_rel_path) -> np.ndarray:
-    image_to_read = img_rel_path
-    image = Image.open(image_to_read)  # [H, W, rgb]
-    image = np.asarray(image)
-    return image
-
-
-def depth_read(filename):
-    depth_in = _read_image(filename)
-    depth_decoded = depth_in / 1000.0
-    return depth_decoded
-
-
-def extract_scannet(
-    root,
-    sample_len=-1,
-    csv_save_path="",
-    datatset_name="",
-    scene_number=16,
-    scene_frames_len=120,
-    stride=1,
-    saved_rgb_dir="",
-    saved_disp_dir="",
-):
-    scenes_names = os.listdir(root)
-    scenes_names = sorted(scenes_names)[:scene_number]
-    all_samples = []
-    for i, seq_name in enumerate(tqdm(scenes_names)):
-        all_img_names = os.listdir(osp.join(root, seq_name, "color"))
-        all_img_names = [x for x in all_img_names if x.endswith(".jpg")]
-        all_img_names = sorted(all_img_names, key=lambda x: int(x.split(".")[0]))
-        all_img_names = all_img_names[:scene_frames_len:stride]
-        print(f"sequence frame number: {len(all_img_names)}")
-
-        seq_len = len(all_img_names)
-        step = sample_len if sample_len > 0 else seq_len
-
-        for ref_idx in range(0, seq_len, step):
-            print(f"Progress: {seq_name}, {ref_idx // step + 1} / {seq_len//step}")
-
-            video_imgs = []
-            video_depths = []
-
-            if (ref_idx + step) <= seq_len:
-                ref_e = ref_idx + step
-            else:
-                continue
-
-            for idx in range(ref_idx, ref_e):
-                im_path = osp.join(root, seq_name, "color", all_img_names[idx])
-                depth_path = osp.join(
-                    root, seq_name, "depth", all_img_names[idx][:-3] + "png"
-                )
-
-                depth = depth_read(depth_path)
-                disp = depth
-
-                video_depths.append(disp)
-                video_imgs.append(np.array(Image.open(im_path)))
-
-            disp_video = np.array(video_depths)[:, None]
-            img_video = np.array(video_imgs)[..., 0:3]
-
-            disp_video = disp_video[:, :, 8:-8, 11:-11]
-            img_video = img_video[:, 8:-8, 11:-11, :]
-
-            data_root = saved_rgb_dir + datatset_name
-            disp_root = saved_disp_dir + datatset_name
-            os.makedirs(data_root, exist_ok=True)
-            os.makedirs(disp_root, exist_ok=True)
-
-            img_video_dir = data_root
-            disp_video_dir = disp_root
-
-            img_video_path = os.path.join(img_video_dir, f"{seq_name}_rgb_left.mp4")
-            disp_video_path = os.path.join(disp_video_dir, f"{seq_name}_disparity.npz")
-
-            imageio.mimsave(
-                img_video_path, img_video, fps=15, quality=9, macro_block_size=1
-            )
-            np.savez(disp_video_path, disparity=disp_video)
-
-            sample = {}
-            sample["filepath_left"] = os.path.join(
-                f"{datatset_name}/{seq_name}_rgb_left.mp4"
-            )
-            sample["filepath_disparity"] = os.path.join(
-                f"{datatset_name}/{seq_name}_disparity.npz"
-            )
-
-            all_samples.append(sample)
-
-    filename_ = csv_save_path
-    os.makedirs(os.path.dirname(filename_), exist_ok=True)
-    fields = ["filepath_left", "filepath_disparity"]
-    with open(filename_, "w") as csvfile:
-        writer = csv.DictWriter(csvfile, fieldnames=fields)
-        writer.writeheader()
-        writer.writerows(all_samples)
-
-    print(f"{filename_} has been saved.")
-
-
-if __name__ == "__main__":
-    extract_scannet(
-        root="path/to/ScanNet_v2/raw/scans_test",
-        saved_rgb_dir="./benchmark/datasets/",
-        saved_disp_dir="./benchmark/datasets/",
-        csv_save_path=f"./benchmark/datasets/scannet.csv",
-        sample_len=-1,
-        datatset_name="scannet",
-        scene_number=100,
-        scene_frames_len=90 * 3,
-        stride=3,
-    )

benchmark/dataset_extract/dataset_extract_sintel.py

@@ -1,137 +0,0 @@
-# Copyright (c) Meta Platforms, Inc. and affiliates.
-# All rights reserved.
-
-# This source code is licensed under the license found in the
-# LICENSE file in the root directory of this source tree.
-# # Data loading based on https://github.com/NVIDIA/flownet2-pytorch
-
-
-import os
-import numpy as np
-import os.path as osp
-from PIL import Image
-from tqdm import tqdm
-import csv
-import imageio
-
-
-# Check for endianness, based on Daniel Scharstein's optical flow code.
-# Using little-endian architecture, these two should be equal.
-TAG_FLOAT = 202021.25
-TAG_CHAR = "PIEH"
-
-
-def depth_read(filename):
-    """Read depth data from file, return as numpy array."""
-    f = open(filename, "rb")
-    check = np.fromfile(f, dtype=np.float32, count=1)[0]
-    assert (
-        check == TAG_FLOAT
-    ), " depth_read:: Wrong tag in flow file (should be: {0}, is: {1}). Big-endian machine? ".format(
-        TAG_FLOAT, check
-    )
-    width = np.fromfile(f, dtype=np.int32, count=1)[0]
-    height = np.fromfile(f, dtype=np.int32, count=1)[0]
-    size = width * height
-    assert (
-        width > 0 and height > 0 and size > 1 and size < 100000000
-    ), " depth_read:: Wrong input size (width = {0}, height = {1}).".format(
-        width, height
-    )
-    depth = np.fromfile(f, dtype=np.float32, count=-1).reshape((height, width))
-    return depth
-
-
-def extract_sintel(
-    root,
-    depth_root,
-    sample_len=-1,
-    csv_save_path="",
-    datatset_name="",
-    saved_rgb_dir="",
-    saved_disp_dir="",
-):
-    scenes_names = os.listdir(root)
-    all_samples = []
-    for i, seq_name in enumerate(tqdm(scenes_names)):
-        all_img_names = os.listdir(os.path.join(root, seq_name))
-        all_img_names = [x for x in all_img_names if x.endswith(".png")]
-        all_img_names.sort()
-        all_img_names = sorted(all_img_names, key=lambda x: int(x.split(".")[0][-4:]))
-
-        seq_len = len(all_img_names)
-        step = sample_len if sample_len > 0 else seq_len
-
-        for ref_idx in range(0, seq_len, step):
-            print(f"Progress: {seq_name}, {ref_idx // step} / {seq_len // step}")
-
-            video_imgs = []
-            video_depths = []
-
-            if (ref_idx + step) <= seq_len:
-                ref_e = ref_idx + step
-            else:
-                continue
-
-            for idx in range(ref_idx, ref_e):
-                im_path = osp.join(root, seq_name, all_img_names[idx])
-                depth_path = osp.join(
-                    depth_root, seq_name, all_img_names[idx][:-3] + "dpt"
-                )
-
-                depth = depth_read(depth_path)
-                disp = depth
-
-                video_depths.append(disp)
-                video_imgs.append(np.array(Image.open(im_path)))
-
-            disp_video = np.array(video_depths)[:, None]
-            img_video = np.array(video_imgs)[..., 0:3]
-
-            data_root = saved_rgb_dir + datatset_name
-            disp_root = saved_disp_dir + datatset_name
-            os.makedirs(data_root, exist_ok=True)
-            os.makedirs(disp_root, exist_ok=True)
-
-            img_video_dir = data_root
-            disp_video_dir = disp_root
-
-            img_video_path = os.path.join(img_video_dir, f"{seq_name}_rgb_left.mp4")
-            disp_video_path = os.path.join(disp_video_dir, f"{seq_name}_disparity.npz")
-
-            imageio.mimsave(
-                img_video_path, img_video, fps=15, quality=10, macro_block_size=1
-            )
-            np.savez(disp_video_path, disparity=disp_video)
-
-            sample = {}
-            sample["filepath_left"] = os.path.join(
-                f"{datatset_name}/{seq_name}_rgb_left.mp4"
-            )
-            sample["filepath_disparity"] = os.path.join(
-                f"{datatset_name}/{seq_name}_disparity.npz"
-            )
-
-            all_samples.append(sample)
-
-    filename_ = csv_save_path
-    os.makedirs(os.path.dirname(filename_), exist_ok=True)
-    fields = ["filepath_left", "filepath_disparity"]
-    with open(filename_, "w") as csvfile:
-        writer = csv.DictWriter(csvfile, fieldnames=fields)
-        writer.writeheader()
-        writer.writerows(all_samples)
-
-    print(f"{filename_} has been saved.")
-
-
-if __name__ == "__main__":
-    extract_sintel(
-        root="path/to/Sintel-Depth/training_image/clean",
-        depth_root="path/to/Sintel-Depth/MPI-Sintel-depth-training-20150305/training/depth",
-        saved_rgb_dir="./benchmark/datasets/",
-        saved_disp_dir="./benchmark/datasets/",
-        csv_save_path=f"./benchmark/datasets/sintel.csv",
-        sample_len=-1,
-        datatset_name="sintel",
-    )

benchmark/dataset_extract_bonn.py

@@ -1,155 +0,0 @@
-import os
-import numpy as np
-import os.path as osp
-from PIL import Image
-from tqdm import tqdm
-import imageio
-import csv
-
-
-def depth_read(filename):
-    # loads depth map D from png file
-    # and returns it as a numpy array
-
-    depth_png = np.asarray(Image.open(filename))
-    # make sure we have a proper 16bit depth map here.. not 8bit!
-    assert np.max(depth_png) > 255
-
-    depth = depth_png.astype(np.float64) / 5000.0
-    depth[depth_png == 0] = -1.0
-    return depth
-
-
-def extract_bonn(
-    root,
-    depth_root,
-    sample_len=-1,
-    csv_save_path="",
-    datatset_name="",
-    saved_rgb_dir="",
-    saved_disp_dir="",
-    start_frame=0,
-    end_frame=110,
-):
-    scenes_names = os.listdir(depth_root)
-    all_samples = []
-    for i, seq_name in enumerate(tqdm(scenes_names)):
-        # load all images
-        all_img_names = os.listdir(osp.join(depth_root, seq_name, "rgb"))
-        all_img_names = [x for x in all_img_names if x.endswith(".png")]
-        print(f"sequence frame number: {len(all_img_names)}")
-
-        # for not zero padding image name
-        all_img_names.sort()
-        all_img_names = sorted(all_img_names, key=lambda x: int(x.split(".")[0][-4:]))
-        all_img_names = all_img_names[start_frame:end_frame]
-
-        all_depth_names = os.listdir(osp.join(depth_root, seq_name, "depth"))
-        all_depth_names = [x for x in all_depth_names if x.endswith(".png")]
-        print(f"sequence depth number: {len(all_depth_names)}")
-
-        # for not zero padding image name
-        all_depth_names.sort()
-        all_depth_names = sorted(
-            all_depth_names, key=lambda x: int(x.split(".")[0][-4:])
-        )
-        all_depth_names = all_depth_names[start_frame:end_frame]
-
-        seq_len = len(all_img_names)
-        step = sample_len if sample_len > 0 else seq_len
-
-        for ref_idx in range(0, seq_len, step):
-            print(f"Progress: {seq_name}, {ref_idx // step + 1} / {seq_len//step}")
-
-            video_imgs = []
-            video_depths = []
-
-            if (ref_idx + step) <= seq_len:
-                ref_e = ref_idx + step
-            else:
-                continue
-
-            # for idx in range(ref_idx, ref_idx + step):
-            for idx in range(ref_idx, ref_e):
-                im_path = osp.join(root, seq_name, "rgb", all_img_names[idx])
-                depth_path = osp.join(
-                    depth_root, seq_name, "depth", all_depth_names[idx]
-                )
-
-                depth = depth_read(depth_path)
-                disp = depth
-
-                video_depths.append(disp)
-                video_imgs.append(np.array(Image.open(im_path)))
-
-            disp_video = np.array(video_depths)[:, None]  # [:, 0:1, :, :, 0]
-            img_video = np.array(video_imgs)[..., 0:3]  # [:, 0, :, :, 0:3]
-
-            print(disp_video.max(), disp_video.min())
-
-            def even_or_odd(num):
-                if num % 2 == 0:
-                    return num
-                else:
-                    return num - 1
-
-            # print(disp_video.shape)
-            # print(img_video.shape)
-            height = disp_video.shape[-2]
-            width = disp_video.shape[-1]
-            height = even_or_odd(height)
-            width = even_or_odd(width)
-            disp_video = disp_video[:, :, 0:height, 0:width]
-            img_video = img_video[:, 0:height, 0:width]
-
-            data_root = saved_rgb_dir + datatset_name
-            disp_root = saved_disp_dir + datatset_name
-            os.makedirs(data_root, exist_ok=True)
-            os.makedirs(disp_root, exist_ok=True)
-
-            img_video_dir = data_root
-            disp_video_dir = disp_root
-
-            img_video_path = os.path.join(img_video_dir, f"{seq_name}_rgb_left.mp4")
-            disp_video_path = os.path.join(disp_video_dir, f"{seq_name}_disparity.npz")
-
-            imageio.mimsave(
-                img_video_path, img_video, fps=15, quality=9, macro_block_size=1
-            )
-            np.savez(disp_video_path, disparity=disp_video)
-
-            sample = {}
-            sample["filepath_left"] = os.path.join(
-                f"{datatset_name}/{seq_name}_rgb_left.mp4"
-            )  # img_video_path
-            sample["filepath_disparity"] = os.path.join(
-                f"{datatset_name}/{seq_name}_disparity.npz"
-            )  # disp_video_path
-
-            all_samples.append(sample)
-
-    # save csv file
-
-    filename_ = csv_save_path
-    os.makedirs(os.path.dirname(filename_), exist_ok=True)
-    fields = ["filepath_left", "filepath_disparity"]
-    with open(filename_, "w") as csvfile:
-        writer = csv.DictWriter(csvfile, fieldnames=fields)
-        writer.writeheader()
-        writer.writerows(all_samples)
-
-    print(f"{filename_} has been saved.")
-
-
-if __name__ == "__main__":
-    extract_bonn(
-        root="path/to/Bonn-RGBD",
-        depth_root="path/to/Bonn-RGBD",
-        saved_rgb_dir="./benchmark/datasets/",
-        saved_disp_dir="./benchmark/datasets/",
-        csv_save_path=f"./benchmark/datasets/bonn.csv",
-        sample_len=-1,
-        datatset_name="bonn",
-        start_frame=30,
-        end_frame=140,
-    )

benchmark/dataset_extract_kitti.py

@@ -1,140 +0,0 @@
-import os
-import numpy as np
-import os.path as osp
-from PIL import Image
-from tqdm import tqdm
-import csv
-import imageio
-
-
-def depth_read(filename):
-    # loads depth map D from png file
-    # and returns it as a numpy array,
-
-    depth_png = np.array(Image.open(filename), dtype=int)
-    # make sure we have a proper 16bit depth map here.. not 8bit!
-    assert np.max(depth_png) > 255
-
-    depth = depth_png.astype(np.float64) / 256.0
-    depth[depth_png == 0] = -1.0
-    return depth
-
-
-def extract_kitti(
-    root,
-    depth_root,
-    sample_len=-1,
-    csv_save_path="",
-    datatset_name="",
-    saved_rgb_dir="",
-    saved_disp_dir="",
-    start_frame=0,
-    end_frame=110,
-):
-    scenes_names = os.listdir(depth_root)
-    all_samples = []
-    for i, seq_name in enumerate(tqdm(scenes_names)):
-        all_img_names = os.listdir(
-            osp.join(depth_root, seq_name, "proj_depth/groundtruth/image_02")
-        )
-        all_img_names = [x for x in all_img_names if x.endswith(".png")]
-        print(f"sequence frame number: {len(all_img_names)}")
-
-        all_img_names.sort()
-        all_img_names = sorted(all_img_names, key=lambda x: int(x.split(".")[0][-4:]))
-        all_img_names = all_img_names[start_frame:end_frame]
-
-        seq_len = len(all_img_names)
-        step = sample_len if sample_len > 0 else seq_len
-
-        for ref_idx in range(0, seq_len, step):
-            print(f"Progress: {seq_name}, {ref_idx // step + 1} / {seq_len//step}")
-
-            video_imgs = []
-            video_depths = []
-
-            if (ref_idx + step) <= seq_len:
-                ref_e = ref_idx + step
-            else:
-                continue
-
-            for idx in range(ref_idx, ref_e):
-                im_path = osp.join(
-                    root, seq_name[0:10], seq_name, "image_02/data", all_img_names[idx]
-                )
-                depth_path = osp.join(
-                    depth_root,
-                    seq_name,
-                    "proj_depth/groundtruth/image_02",
-                    all_img_names[idx],
-                )
-
-                depth = depth_read(depth_path)
-                disp = depth
-
-                video_depths.append(disp)
-                video_imgs.append(np.array(Image.open(im_path)))
-
-            disp_video = np.array(video_depths)[:, None]
-            img_video = np.array(video_imgs)[..., 0:3]
-
-            def even_or_odd(num):
-                if num % 2 == 0:
-                    return num
-                else:
-                    return num - 1
-
-            height = disp_video.shape[-2]
-            width = disp_video.shape[-1]
-            height = even_or_odd(height)
-            width = even_or_odd(width)
-            disp_video = disp_video[:, :, 0:height, 0:width]
-            img_video = img_video[:, 0:height, 0:width]
-
-            data_root = saved_rgb_dir + datatset_name
-            disp_root = saved_disp_dir + datatset_name
-            os.makedirs(data_root, exist_ok=True)
-            os.makedirs(disp_root, exist_ok=True)
-
-            img_video_dir = data_root
-            disp_video_dir = disp_root
-
-            img_video_path = os.path.join(img_video_dir, f"{seq_name}_rgb_left.mp4")
-            disp_video_path = os.path.join(disp_video_dir, f"{seq_name}_disparity.npz")
-
-            imageio.mimsave(
-                img_video_path, img_video, fps=15, quality=10, macro_block_size=1
-            )
-            np.savez(disp_video_path, disparity=disp_video)
-
-            sample = {}
-            sample["filepath_left"] = os.path.join(f"KITTI/{seq_name}_rgb_left.mp4")
-            sample["filepath_disparity"] = os.path.join(
-                f"KITTI/{seq_name}_disparity.npz"
-            )
-
-            all_samples.append(sample)
-
-    filename_ = csv_save_path
-    os.makedirs(os.path.dirname(filename_), exist_ok=True)
-    fields = ["filepath_left", "filepath_disparity"]
-    with open(filename_, "w") as csvfile:
-        writer = csv.DictWriter(csvfile, fieldnames=fields)
-        writer.writeheader()
-        writer.writerows(all_samples)
-
-    print(f"{filename_} has been saved.")
-
-
-if __name__ == "__main__":
-    extract_kitti(
-        root="path/to/KITTI/raw_data",
-        depth_root="path/to/KITTI/data_depth_annotated/val",
-        saved_rgb_dir="./benchmark/datasets/",
-        saved_disp_dir="./benchmark/datasets/",
-        csv_save_path=f"./benchmark/datasets/KITTI.csv",
-        sample_len=-1,
-        datatset_name="KITTI",
-        start_frame=0,
-        end_frame=110,
-    )

benchmark/dataset_extract_nyu.py

@@ -1,106 +0,0 @@
-import os
-import numpy as np
-import os.path as osp
-from PIL import Image
-from tqdm import tqdm
-import csv
-import imageio
-
-
-def _read_image(img_rel_path) -> np.ndarray:
-    image_to_read = img_rel_path
-    image = Image.open(image_to_read)
-    image = np.asarray(image)
-    return image
-
-
-def depth_read(filename):
-    depth_in = _read_image(filename)
-    depth_decoded = depth_in / 1000.0
-    return depth_decoded
-
-
-def extract_nyu(
-    root,
-    depth_root,
-    csv_save_path="",
-    datatset_name="",
-    filename_ls_path="",
-    saved_rgb_dir="",
-    saved_disp_dir="",
-):
-    with open(filename_ls_path, "r") as f:
-        filenames = [s.split() for s in f.readlines()]
-
-    all_samples = []
-    for i, pair_names in enumerate(tqdm(filenames)):
-        img_name = pair_names[0]
-        filled_depth_name = pair_names[2]
-
-        im_path = osp.join(root, img_name)
-        depth_path = osp.join(depth_root, filled_depth_name)
-
-        depth = depth_read(depth_path)
-        disp = depth
-
-        video_depths = [disp]
-        video_imgs = [np.array(Image.open(im_path))]
-
-        disp_video = np.array(video_depths)[:, None]
-        img_video = np.array(video_imgs)[..., 0:3]
-
-        disp_video = disp_video[:, :, 45:471, 41:601]
-        img_video = img_video[:, 45:471, 41:601, :]
-
-        data_root = saved_rgb_dir + datatset_name
-        disp_root = saved_disp_dir + datatset_name
-        os.makedirs(data_root, exist_ok=True)
-        os.makedirs(disp_root, exist_ok=True)
-
-        img_video_dir = data_root
-        disp_video_dir = disp_root
-
-        img_video_path = os.path.join(img_video_dir, f"{img_name[:-4]}_rgb_left.mp4")
-        disp_video_path = os.path.join(disp_video_dir, f"{img_name[:-4]}_disparity.npz")
-
-        dir_name = os.path.dirname(img_video_path)
-        os.makedirs(dir_name, exist_ok=True)
-        dir_name = os.path.dirname(disp_video_path)
-        os.makedirs(dir_name, exist_ok=True)
-
-        imageio.mimsave(
-            img_video_path, img_video, fps=15, quality=10, macro_block_size=1
-        )
-        np.savez(disp_video_path, disparity=disp_video)
-
-        sample = {}
-        sample["filepath_left"] = os.path.join(
-            f"{datatset_name}/{img_name[:-4]}_rgb_left.mp4"
-        )
-        sample["filepath_disparity"] = os.path.join(
-            f"{datatset_name}/{img_name[:-4]}_disparity.npz"
-        )
-
-        all_samples.append(sample)
-
-    filename_ = csv_save_path
-    os.makedirs(os.path.dirname(filename_), exist_ok=True)
-    fields = ["filepath_left", "filepath_disparity"]
-    with open(filename_, "w") as csvfile:
-        writer = csv.DictWriter(csvfile, fieldnames=fields)
-        writer.writeheader()
-        writer.writerows(all_samples)
-
-    print(f"{filename_} has been saved.")
-
-
-if __name__ == "__main__":
-    extract_nyu(
-        root="path/to/NYUv2/",
-        depth_root="path/to/NYUv2/",
-        filename_ls_path="path/to/NYUv2/filename_list_test.txt",
-        saved_rgb_dir="./benchmark/datasets/",
-        saved_disp_dir="./benchmark/datasets/",
-        csv_save_path=f"./benchmark/datasets/NYUv2.csv",
-        datatset_name="NYUv2",
-    )

benchmark/dataset_extract_scannet.py

@@ -1,124 +0,0 @@
-import os
-import numpy as np
-import os.path as osp
-from PIL import Image
-from tqdm import tqdm
-import csv
-import imageio
-
-
-def _read_image(img_rel_path) -> np.ndarray:
-    image_to_read = img_rel_path
-    image = Image.open(image_to_read)  # [H, W, rgb]
-    image = np.asarray(image)
-    return image
-
-
-def depth_read(filename):
-    depth_in = _read_image(filename)
-    depth_decoded = depth_in / 1000.0
-    return depth_decoded
-
-
-def extract_scannet(
-    root,
-    sample_len=-1,
-    csv_save_path="",
-    datatset_name="",
-    scene_number=16,
-    scene_frames_len=120,
-    stride=1,
-    saved_rgb_dir="",
-    saved_disp_dir="",
-):
-    scenes_names = os.listdir(root)
-    scenes_names = sorted(scenes_names)[:scene_number]
-    all_samples = []
-    for i, seq_name in enumerate(tqdm(scenes_names)):
-        all_img_names = os.listdir(osp.join(root, seq_name, "color"))
-        all_img_names = [x for x in all_img_names if x.endswith(".jpg")]
-        all_img_names = sorted(all_img_names, key=lambda x: int(x.split(".")[0]))
-        all_img_names = all_img_names[:scene_frames_len:stride]
-        print(f"sequence frame number: {len(all_img_names)}")
-
-        seq_len = len(all_img_names)
-        step = sample_len if sample_len > 0 else seq_len
-
-        for ref_idx in range(0, seq_len, step):
-            print(f"Progress: {seq_name}, {ref_idx // step + 1} / {seq_len//step}")
-
-            video_imgs = []
-            video_depths = []
-
-            if (ref_idx + step) <= seq_len:
-                ref_e = ref_idx + step
-            else:
-                continue
-
-            for idx in range(ref_idx, ref_e):
-                im_path = osp.join(root, seq_name, "color", all_img_names[idx])
-                depth_path = osp.join(
-                    root, seq_name, "depth", all_img_names[idx][:-3] + "png"
-                )
-
-                depth = depth_read(depth_path)
-                disp = depth
-
-                video_depths.append(disp)
-                video_imgs.append(np.array(Image.open(im_path)))
-
-            disp_video = np.array(video_depths)[:, None]
-            img_video = np.array(video_imgs)[..., 0:3]
-
-            disp_video = disp_video[:, :, 8:-8, 11:-11]
-            img_video = img_video[:, 8:-8, 11:-11, :]
-
-            data_root = saved_rgb_dir + datatset_name
-            disp_root = saved_disp_dir + datatset_name
-            os.makedirs(data_root, exist_ok=True)
-            os.makedirs(disp_root, exist_ok=True)
-
-            img_video_dir = data_root
-            disp_video_dir = disp_root
-
-            img_video_path = os.path.join(img_video_dir, f"{seq_name}_rgb_left.mp4")
-            disp_video_path = os.path.join(disp_video_dir, f"{seq_name}_disparity.npz")
-
-            imageio.mimsave(
-                img_video_path, img_video, fps=15, quality=9, macro_block_size=1
-            )
-            np.savez(disp_video_path, disparity=disp_video)
-
-            sample = {}
-            sample["filepath_left"] = os.path.join(
-                f"{datatset_name}/{seq_name}_rgb_left.mp4"
-            )
-            sample["filepath_disparity"] = os.path.join(
-                f"{datatset_name}/{seq_name}_disparity.npz"
-            )
-
-            all_samples.append(sample)
-
-    filename_ = csv_save_path
-    os.makedirs(os.path.dirname(filename_), exist_ok=True)
-    fields = ["filepath_left", "filepath_disparity"]
-    with open(filename_, "w") as csvfile:
-        writer = csv.DictWriter(csvfile, fieldnames=fields)
-        writer.writeheader()
-        writer.writerows(all_samples)
-
-    print(f"{filename_} has been saved.")
-
-
-if __name__ == "__main__":
-    extract_scannet(
-        root="path/to/ScanNet_v2/raw/scans_test",
-        saved_rgb_dir="./benchmark/datasets/",
-        saved_disp_dir="./benchmark/datasets/",
-        csv_save_path=f"./benchmark/datasets/scannet.csv",
-        sample_len=-1,
-        datatset_name="scannet",
-        scene_number=100,
-        scene_frames_len=90 * 3,
-        stride=3,
-    )

benchmark/dataset_extract_sintel.py

@@ -1,137 +0,0 @@
-# Copyright (c) Meta Platforms, Inc. and affiliates.
-# All rights reserved.
-
-# This source code is licensed under the license found in the
-# LICENSE file in the root directory of this source tree.
-# # Data loading based on https://github.com/NVIDIA/flownet2-pytorch
-
-
-import os
-import numpy as np
-import os.path as osp
-from PIL import Image
-from tqdm import tqdm
-import csv
-import imageio
-
-
-# Check for endianness, based on Daniel Scharstein's optical flow code.
-# Using little-endian architecture, these two should be equal.
-TAG_FLOAT = 202021.25
-TAG_CHAR = "PIEH"
-
-
-def depth_read(filename):
-    """Read depth data from file, return as numpy array."""
-    f = open(filename, "rb")
-    check = np.fromfile(f, dtype=np.float32, count=1)[0]
-    assert (
-        check == TAG_FLOAT
-    ), " depth_read:: Wrong tag in flow file (should be: {0}, is: {1}). Big-endian machine? ".format(
-        TAG_FLOAT, check
-    )
-    width = np.fromfile(f, dtype=np.int32, count=1)[0]
-    height = np.fromfile(f, dtype=np.int32, count=1)[0]
-    size = width * height
-    assert (
-        width > 0 and height > 0 and size > 1 and size < 100000000
-    ), " depth_read:: Wrong input size (width = {0}, height = {1}).".format(
-        width, height
-    )
-    depth = np.fromfile(f, dtype=np.float32, count=-1).reshape((height, width))
-    return depth
-
-
-def extract_sintel(
-    root,
-    depth_root,
-    sample_len=-1,
-    csv_save_path="",
-    datatset_name="",
-    saved_rgb_dir="",
-    saved_disp_dir="",
-):
-    scenes_names = os.listdir(root)
-    all_samples = []
-    for i, seq_name in enumerate(tqdm(scenes_names)):
-        all_img_names = os.listdir(os.path.join(root, seq_name))
-        all_img_names = [x for x in all_img_names if x.endswith(".png")]
-        all_img_names.sort()
-        all_img_names = sorted(all_img_names, key=lambda x: int(x.split(".")[0][-4:]))
-
-        seq_len = len(all_img_names)
-        step = sample_len if sample_len > 0 else seq_len
-
-        for ref_idx in range(0, seq_len, step):
-            print(f"Progress: {seq_name}, {ref_idx // step} / {seq_len // step}")
-
-            video_imgs = []
-            video_depths = []
-
-            if (ref_idx + step) <= seq_len:
-                ref_e = ref_idx + step
-            else:
-                continue
-
-            for idx in range(ref_idx, ref_e):
-                im_path = osp.join(root, seq_name, all_img_names[idx])
-                depth_path = osp.join(
-                    depth_root, seq_name, all_img_names[idx][:-3] + "dpt"
-                )
-
-                depth = depth_read(depth_path)
-                disp = depth
-
-                video_depths.append(disp)
-                video_imgs.append(np.array(Image.open(im_path)))
-
-            disp_video = np.array(video_depths)[:, None]
-            img_video = np.array(video_imgs)[..., 0:3]
-
-            data_root = saved_rgb_dir + datatset_name
-            disp_root = saved_disp_dir + datatset_name
-            os.makedirs(data_root, exist_ok=True)
-            os.makedirs(disp_root, exist_ok=True)
-
-            img_video_dir = data_root
-            disp_video_dir = disp_root
-
-            img_video_path = os.path.join(img_video_dir, f"{seq_name}_rgb_left.mp4")
-            disp_video_path = os.path.join(disp_video_dir, f"{seq_name}_disparity.npz")
-
-            imageio.mimsave(
-                img_video_path, img_video, fps=15, quality=10, macro_block_size=1
-            )
-            np.savez(disp_video_path, disparity=disp_video)
-
-            sample = {}
-            sample["filepath_left"] = os.path.join(
-                f"{datatset_name}/{seq_name}_rgb_left.mp4"
-            )
-            sample["filepath_disparity"] = os.path.join(
-                f"{datatset_name}/{seq_name}_disparity.npz"
-            )
-
-            all_samples.append(sample)
-
-    filename_ = csv_save_path
-    os.makedirs(os.path.dirname(filename_), exist_ok=True)
-    fields = ["filepath_left", "filepath_disparity"]
-    with open(filename_, "w") as csvfile:
-        writer = csv.DictWriter(csvfile, fieldnames=fields)
-        writer.writeheader()
-        writer.writerows(all_samples)
-
-    print(f"{filename_} has been saved.")
-
-
-if __name__ == "__main__":
-    extract_sintel(
-        root="path/to/Sintel-Depth/training_image/clean",
-        depth_root="path/to/Sintel-Depth/MPI-Sintel-depth-training-20150305/training/depth",
-        saved_rgb_dir="./benchmark/datasets/",
-        saved_disp_dir="./benchmark/datasets/",
-        csv_save_path=f"./benchmark/datasets/sintel.csv",
-        sample_len=-1,
-        datatset_name="sintel",
-    )

benchmark/demo.sh

@@ -1,18 +0,0 @@
-#!/bin/sh
-set -x
-set -e
-
-test_case=$1
-gpu_id=$2
-process_length=$3
-saved_root=$4
-saved_dataset_folder=$5
-overlap=$6
-dataset=$7
-
-CUDA_VISIBLE_DEVICES=${gpu_id} PYTHONPATH=. python run.py \
-  --video-path ${test_case} \
-  --save-folder ${saved_root}/${saved_dataset_folder} \
-  --process-length ${process_length} \
-  --dataset ${dataset} \
-  --overlap ${overlap}

benchmark/eval/eval.py

@@ -1,282 +0,0 @@
-import numpy as np
-import os
-import torch
-import cv2
-import csv
-from metric import * 
-import metric
-import argparse
-from tqdm import tqdm
-import json
-
-
-device = 'cuda'
-eval_metrics = [
-    "abs_relative_difference",
-    "rmse_linear",
-    "delta1_acc",
-    # "squared_relative_difference",
-    # "rmse_log",
-    # "log10",
-    # "delta2_acc",
-    # "delta3_acc",
-    # "i_rmse",
-    # "silog_rmse",
-]
-
-
-def depth2disparity(depth, return_mask=False):
-    if isinstance(depth, torch.Tensor):
-        disparity = torch.zeros_like(depth)
-    elif isinstance(depth, np.ndarray):
-        disparity = np.zeros_like(depth)
-    non_negtive_mask = depth > 0
-    disparity[non_negtive_mask] = 1.0 / depth[non_negtive_mask]
-    if return_mask:
-        return disparity, non_negtive_mask
-    else:
-        return disparity
-
-
-def resize_images(images, new_size):
-    resized_images = np.empty(
-        (images.shape[0], new_size[0], new_size[1], images.shape[3])
-    )
-
-    for i, image in enumerate(images):
-        if image.shape[2]==1:
-            resized_images[i] = cv2.resize(image, (new_size[1], new_size[0]))[..., None]
-        else:
-            resized_images[i] = cv2.resize(image, (new_size[1], new_size[0]))
-
-    return resized_images
-    
-
-def eval_single(
-    pred_disp_path, 
-    gt_disp_path, 
-    seq_len=98, 
-    domain='depth', 
-    method_type="ours",
-    dataset_max_depth="70"
-):
-    # load data
-    gt_disp = np.load(gt_disp_path)['disparity'] \
-        if 'disparity' in np.load(gt_disp_path).files else \
-        np.load(gt_disp_path)['arr_0']  # (t, 1, h, w)
-
-    if method_type=="ours":
-        pred_disp = np.load(pred_disp_path)['depth'] # (t, h, w)
-    if method_type=="depth_anything":
-        pred_disp = np.load(pred_disp_path)['disparity'] # (t, h, w)
-    
-    # seq_len
-    if pred_disp.shape[0] < seq_len:
-         seq_len = pred_disp.shape[0]
-
-    # preprocess
-    pred_disp = resize_images(pred_disp[..., None], (gt_disp.shape[-2], gt_disp.shape[-1])) # (t, h, w)
-    pred_disp = pred_disp[..., 0] # (t, h, w)
-    pred_disp = pred_disp[:seq_len]
-    gt_disp = gt_disp[:seq_len, 0] # (t, h, w)
-
-    # valid mask
-    valid_mask = np.logical_and(
-            (gt_disp > 1e-3), 
-            (gt_disp < dataset_max_depth)
-        )
-    pred_disp = np.clip(pred_disp, a_min=1e-3, a_max=None) 
-    pred_disp_masked = pred_disp[valid_mask].reshape((-1, 1))
-    
-    # choose evaluation domain
-    DOMAIN = domain
-    if DOMAIN=='disp':
-        # align in real disp, calc in disp
-        gt_disp_maksed = gt_disp[valid_mask].reshape((-1, 1)).astype(np.float64)
-    elif DOMAIN=='depth':
-        # align in disp = 1/depth, calc in depth
-        gt_disp_maksed = 1. / (gt_disp[valid_mask].reshape((-1, 1)).astype(np.float64) + 1e-8)
-    else:
-        pass
-
-
-    # calc scale and shift
-    _ones = np.ones_like(pred_disp_masked)
-    A = np.concatenate([pred_disp_masked, _ones], axis=-1)
-    X = np.linalg.lstsq(A, gt_disp_maksed, rcond=None)[0]
-    scale, shift = X # gt = scale * pred + shift
-    
-    # align
-    aligned_pred = scale * pred_disp + shift
-    aligned_pred = np.clip(aligned_pred, a_min=1e-3, a_max=None) 
-
-
-    # align in real disp, calc in disp
-    if DOMAIN=='disp':
-        pred_depth = aligned_pred
-        gt_depth = gt_disp
-    # align in disp = 1/depth, calc in depth
-    elif DOMAIN=='depth':
-        pred_depth = depth2disparity(aligned_pred)
-        gt_depth = gt_disp
-    else:
-        pass
-
-    # metric evaluation, clip to dataset min max
-    pred_depth = np.clip(
-            pred_depth, a_min=1e-3, a_max=dataset_max_depth
-        )
-
-    # evaluate metric 
-    sample_metric = []
-    metric_funcs = [getattr(metric, _met) for _met in eval_metrics]
-
-    # Evaluate 
-    sample_metric = []
-    pred_depth_ts = torch.from_numpy(pred_depth).to(device)
-    gt_depth_ts = torch.from_numpy(gt_depth).to(device)
-    valid_mask_ts = torch.from_numpy(valid_mask).to(device)
-
-    n = valid_mask.sum((-1, -2))
-    valid_frame = (n > 0)
-    pred_depth_ts = pred_depth_ts[valid_frame]
-    gt_depth_ts = gt_depth_ts[valid_frame]
-    valid_mask_ts = valid_mask_ts[valid_frame]
-
-    for met_func in metric_funcs:
-        _metric_name = met_func.__name__
-        _metric = met_func(pred_depth_ts, gt_depth_ts, valid_mask_ts).item()
-        sample_metric.append(_metric)
-
-    return sample_metric
-
-
-
-if __name__=="__main__":
-    parser = argparse.ArgumentParser()
-
-    parser.add_argument(
-        "--seq_len", 
-        type=int,
-        default=50,
-        help="Max video frame length for evaluation."
-    )
-
-    parser.add_argument(
-        "--domain",
-        type=str,
-        default="depth",
-        choices=["depth", "disp"],
-        help="Domain of metric calculation."
-    )
-
-    parser.add_argument(
-        "--method_type",
-        type=str,
-        default="ours",
-        choices=["ours", "depth_anything"],
-        help="Choose the methods."
-    )
-
-    parser.add_argument(
-        "--dataset_max_depth",
-        type=int,
-        default=70,
-        help="Dataset max depth clip."
-    )
-
-    parser.add_argument(
-        "--pred_disp_root",
-        type=str,
-        default="./demo_output",
-        help="Predicted output directory."
-    )
-
-    parser.add_argument(
-        "--gt_disp_root",
-        type=str,
-        required=True,
-        help="GT depth directory."
-    )
-
-    parser.add_argument(
-        "--dataset",
-        type=str,
-        required=True,
-        help="Choose the datasets."
-    )
-
-    parser.add_argument(
-        "--meta_path",
-        type=str,
-        required=True,
-        help="Path of test dataset csv file."
-    )
-
-
-    args = parser.parse_args()
-
-    SEQ_LEN = args.seq_len
-    method_type = args.method_type
-    if method_type == "ours":
-        pred_disp_root = os.path.join(args.pred_disp_root, f'results_{args.dataset}')
-    else:
-        # pred_disp_root = args.pred_disp_root
-        pred_disp_root = os.path.join(args.pred_disp_root, f'results_{args.dataset}')
-    domain = args.domain
-    dataset_max_depth = args.dataset_max_depth
-    saved_json_path = os.path.join(args.pred_disp_root, f"results_{args.dataset}.json")
-
-    meta_path = args.meta_path
-
-    assert method_type in ["depth_anything", "ours"], "Invalid method type, must be in ['depth_anything', 'ours']"
-    assert domain in ["depth", "disp"], "Invalid domain type, must be in ['depth', 'disp']"
-        
-    with open(meta_path, mode="r", encoding="utf-8") as csvfile:
-        csv_reader = csv.DictReader(csvfile)
-        samples = list(csv_reader)
-
-    # iterate all cases
-    results_all = []
-    for i, sample in enumerate(tqdm(samples)):
-        gt_disp_path = os.path.join(args.gt_disp_root, samples[i]['filepath_disparity'])
-        if method_type=="ours":
-            pred_disp_path = os.path.join(pred_disp_root, samples[i]['filepath_disparity'])
-            pred_disp_path = pred_disp_path.replace("disparity", "rgb_left")
-        
-        if method_type=="depth_anything":
-            pred_disp_path = os.path.join(pred_disp_root, samples[i]['filepath_disparity'])
-            pred_disp_path = pred_disp_path.replace("disparity", "rgb_left_depth")
-        
-        results_single = eval_single(
-            pred_disp_path, 
-            gt_disp_path, 
-            seq_len=SEQ_LEN, 
-            domain=domain, 
-            method_type=method_type, 
-            dataset_max_depth=dataset_max_depth
-        )
-
-        results_all.append(results_single)
-
-    # avarage
-    final_results =  np.array(results_all)
-    final_results_mean = np.mean(final_results, axis=0)
-    print("")
-
-    # save mean to json
-    result_dict = { 'name': method_type }
-    for i, metric in enumerate(eval_metrics):
-        result_dict[metric] = final_results_mean[i]
-        print(f"{metric}: {final_results_mean[i]:04f}")
-
-    # save each case to json
-    for i, results in enumerate(results_all):
-        result_dict[samples[i]['filepath_disparity']] = results
-
-    # write json
-    with open(saved_json_path, 'w') as f:
-        json.dump(result_dict, f, indent=4)
-    print("")
-    print(f"Evaluation results json are saved to {saved_json_path}")
-    

benchmark/eval/eval.sh

@@ -1,51 +0,0 @@
-#!/bin/sh
-set -x
-set -e
-
-pred_disp_root=/path/to/saved/root_directory # The parent directory that contaning [sintel, scannet, KITTI, bonn, NYUv2] prediction
-gt_disp_root=/path/to/gt_depth/root_directory # The parent directory that contaning [sintel, scannet, KITTI, bonn, NYUv2] ground truth
-
-# eval sintel
-python benchmark/eval/eval.py \
-    --meta_path ./eval/csv/meta_sintel.csv \
-    --dataset_max_depth 70 \
-    --dataset sintel \
-    --seq_len 50 \
-    --pred_disp_root ${pred_disp_root} \
-    --gt_disp_root ${gt_disp_root} \
-
-# eval scannet
-python benchmark/eval/eval.py \
-    --meta_path ./eval/csv/meta_scannet_test.csv \
-    --dataset_max_depth 10 \
-    --dataset scannet \
-    --seq_len 90 \
-    --pred_disp_root ${pred_disp_root} \
-    --gt_disp_root ${gt_disp_root} \
-
-# eval kitti
-python benchmark/eval/eval.py \
-    --meta_path ./eval/csv/meta_kitti_val.csv \
-    --dataset_max_depth 80 \
-    --dataset kitti \
-    --seq_len 110 \
-    --pred_disp_root ${pred_disp_root} \
-    --gt_disp_root ${gt_disp_root} \
-
-# eval bonn
-python benchmark/eval/eval.py \
-    --meta_path ./eval/csv/meta_bonn.csv \
-    --dataset_max_depth 10 \
-    --dataset bonn \
-    --seq_len 110 \
-    --pred_disp_root ${pred_disp_root} \
-    --gt_disp_root ${gt_disp_root} \
-
-# eval nyu
-python benchmark/eval/eval.py \
-    --meta_path ./eval/csv/meta_nyu_test.csv \
-    --dataset_max_depth 10 \
-    --dataset nyu \
-    --seq_len 1 \
-    --pred_disp_root ${pred_disp_root} \
-    --gt_disp_root ${gt_disp_root} \

benchmark/eval/metric.py

@@ -1,128 +0,0 @@
-import torch
-
-
-def abs_relative_difference(output, target, valid_mask=None):
-    actual_output = output
-    actual_target = target
-    abs_relative_diff = torch.abs(actual_output - actual_target) / actual_target
-    if valid_mask is not None:
-        abs_relative_diff[~valid_mask] = 0
-        n = valid_mask.sum((-1, -2))
-    else:
-        n = output.shape[-1] * output.shape[-2]
-    abs_relative_diff = torch.sum(abs_relative_diff, (-1, -2)) / n
-    return abs_relative_diff.mean()
-
-
-def squared_relative_difference(output, target, valid_mask=None):
-    actual_output = output
-    actual_target = target
-    square_relative_diff = (
-        torch.pow(torch.abs(actual_output - actual_target), 2) / actual_target
-    )
-    if valid_mask is not None:
-        square_relative_diff[~valid_mask] = 0
-        n = valid_mask.sum((-1, -2))
-    else:
-        n = output.shape[-1] * output.shape[-2]
-    square_relative_diff = torch.sum(square_relative_diff, (-1, -2)) / n
-    return square_relative_diff.mean()
-
-
-def rmse_linear(output, target, valid_mask=None):
-    actual_output = output
-    actual_target = target
-    diff = actual_output - actual_target
-    if valid_mask is not None:
-        diff[~valid_mask] = 0
-        n = valid_mask.sum((-1, -2))
-    else:
-        n = output.shape[-1] * output.shape[-2]
-    diff2 = torch.pow(diff, 2)
-    mse = torch.sum(diff2, (-1, -2)) / n
-    rmse = torch.sqrt(mse)
-    return rmse.mean()
-
-
-def rmse_log(output, target, valid_mask=None):
-    diff = torch.log(output) - torch.log(target)
-    if valid_mask is not None:
-        diff[~valid_mask] = 0
-        n = valid_mask.sum((-1, -2))
-    else:
-        n = output.shape[-1] * output.shape[-2]
-    diff2 = torch.pow(diff, 2)
-    mse = torch.sum(diff2, (-1, -2)) / n  # [B]
-    rmse = torch.sqrt(mse)
-    return rmse.mean()
-
-
-def log10(output, target, valid_mask=None):
-    if valid_mask is not None:
-        diff = torch.abs(
-            torch.log10(output[valid_mask]) - torch.log10(target[valid_mask])
-        )
-    else:
-        diff = torch.abs(torch.log10(output) - torch.log10(target))
-    return diff.mean()
-
-
-# adapt from: https://github.com/imran3180/depth-map-prediction/blob/master/main.py
-def threshold_percentage(output, target, threshold_val, valid_mask=None):
-    d1 = output / target
-    d2 = target / output
-    max_d1_d2 = torch.max(d1, d2)
-    zero = torch.zeros(*output.shape)
-    one = torch.ones(*output.shape)
-    bit_mat = torch.where(max_d1_d2.cpu() < threshold_val, one, zero)
-    if valid_mask is not None:
-        bit_mat[~valid_mask] = 0
-        n = valid_mask.sum((-1, -2))
-    else:
-        n = output.shape[-1] * output.shape[-2]
-    count_mat = torch.sum(bit_mat, (-1, -2))
-    threshold_mat = count_mat / n.cpu()
-    return threshold_mat.mean()
-
-
-def delta1_acc(pred, gt, valid_mask):
-    return threshold_percentage(pred, gt, 1.25, valid_mask)
-
-
-def delta2_acc(pred, gt, valid_mask):
-    return threshold_percentage(pred, gt, 1.25**2, valid_mask)
-
-
-def delta3_acc(pred, gt, valid_mask):
-    return threshold_percentage(pred, gt, 1.25**3, valid_mask)
-
-
-def i_rmse(output, target, valid_mask=None):
-    output_inv = 1.0 / output
-    target_inv = 1.0 / target
-    diff = output_inv - target_inv
-    if valid_mask is not None:
-        diff[~valid_mask] = 0
-        n = valid_mask.sum((-1, -2))
-    else:
-        n = output.shape[-1] * output.shape[-2]
-    diff2 = torch.pow(diff, 2)
-    mse = torch.sum(diff2, (-1, -2)) / n  # [B]
-    rmse = torch.sqrt(mse)
-    return rmse.mean()
-
-
-def silog_rmse(depth_pred, depth_gt, valid_mask=None):
-    diff = torch.log(depth_pred) - torch.log(depth_gt)
-    if valid_mask is not None:
-        diff[~valid_mask] = 0
-        n = valid_mask.sum((-1, -2))
-    else:
-        n = depth_gt.shape[-2] * depth_gt.shape[-1]
-
-    diff2 = torch.pow(diff, 2)
-
-    first_term = torch.sum(diff2, (-1, -2)) / n
-    second_term = torch.pow(torch.sum(diff, (-1, -2)), 2) / (n**2)
-    loss = torch.sqrt(torch.mean(first_term - second_term)) * 100
-    return loss

benchmark/infer/infer.sh

@@ -1,55 +0,0 @@
-#!/bin/sh
-set -x
-set -e
-
-input_rgb_root=/path/to/input/RGB/root_directory # The parent directory that contaning [sintel, scannet, KITTI, bonn, NYUv2] input RGB
-saved_root=/path/to/saved/root_directory # The parent directory that saving [sintel, scannet, KITTI, bonn, NYUv2] prediction
-gpus=0,1,2,3 # Using 4 GPU, you can adjust it according to your device
-
-
-# infer sintel
-python benchmark/infer/infer_batch.py \
-    --meta_path ./eval/csv/meta_sintel.csv \
-    --saved_root ${saved_root} \
-    --saved_dataset_folder results_sintel \
-    --process_length 50 \
-    --gpus ${gpus} \
-    --dataset sintel \
-
-# infer scannet
-python benchmark/infer/infer_batch.py \
-    --meta_path ./eval/csv/meta_scannet_test.csv \
-    --saved_root ${saved_root} \
-    --saved_dataset_folder results_scannet \
-    --process_length 90 \
-    --gpus ${gpus} \
-    --dataset scannet \
-
-# infer kitti
-python benchmark/infer/infer_batch.py \
-    --meta_path ./eval/csv/meta_kitti_val.csv \
-    --saved_root ${saved_root} \
-    --saved_dataset_folder results_kitti \
-    --process_length 110 \
-    --gpus ${gpus} \
-    --dataset kitti \
-
-# infer bonn
-python benchmark/infer/infer_batch.py \
-    --meta_path ./eval/csv/meta_bonn.csv \
-    --saved_root ${saved_root} \
-    --saved_dataset_folder results_bonn \
-    --input_rgb_root ${input_rgb_root} \
-    --process_length 110 \
-    --gpus ${gpus} \
-    --dataset bonn \
-
-# infer nyu
-python benchmark/infer/infer_batch.py \
-    --meta_path ./eval/csv/meta_nyu_test.csv \
-    --saved_root ${saved_root} \
-    --saved_dataset_folder results_nyu \
-    --process_length 1 \
-    --gpus ${gpus} \
-    --overlap 0 \
-    --dataset nyu \

benchmark/infer/infer_batch.py

@@ -1,46 +0,0 @@
-import os
-import multiprocessing as mp
-import csv
-import argparse
-
-
-def process_video(video_path, gpu_id, save_folder, args):
-    os.system(f'sh ./benchmark/demo.sh {video_path} {gpu_id} {int(args.process_length)} {args.saved_root} {save_folder} {args.overlap} {args.dataset}')
-
-if __name__ == '__main__':
-
-    parser = argparse.ArgumentParser()
-    
-    parser.add_argument('--meta_path', type=str)
-    parser.add_argument('--saved_dataset_folder', type=str)
-    parser.add_argument('--saved_root', type=str, default="./output")
-    parser.add_argument('--input_rgb_root', type=str)
-
-    parser.add_argument('--process_length', type=int, default=110)
-    parser.add_argument('--gpus', type=str, default="0,1,2,3")
-    
-    parser.add_argument('--overlap', type=int, default=1)
-    parser.add_argument('--dataset', type=str, default="open")
-    
-    args = parser.parse_args()
-    gpus = args.gpus.strip().split(',')
-
-    with open(args.meta_path, mode="r", encoding="utf-8") as csvfile:
-        csv_reader = csv.DictReader(csvfile)
-        test_samples = list(csv_reader)
-    batch_size = len(gpus)
-    video_batches = [test_samples[i:i+batch_size] for i in range(0, len(test_samples), batch_size)]
-    print("gpus+++: ", gpus)
-
-    processes = []
-    for video_batch in video_batches:
-        for i, sample in enumerate(video_batch):
-            video_path = os.path.join(args.input_rgb_root, sample["filepath_left"])
-            save_folder = os.path.join(args.saved_dataset_folder, os.path.dirname(sample["filepath_left"]))
-            gpu_id = gpus[i % len(gpus)]
-            p = mp.Process(target=process_video, args=(video_path, gpu_id, save_folder, args))
-            p.start()
-            processes.append(p)
-        
-        for p in processes: 
-            p.join()