Update app.py

app.py

@@ -36,7 +36,8 @@ def stitch_rgbd_videos(
         # For stitching: read the original video in full resolution (without downscaling).
         full_frames, target_fps = read_video_frames(processed_video, max_len, target_fps, max_res=-1)
         depths, _ = read_video_frames(depth_vis_video, max_len, target_fps, max_res=-1)
-        
+
+        print(f"Depth frame shape: {depths[0].shape}, dtype: {depths[0].dtype}, min: {depths[0].min()}, max: {depths[0].max()}")
         
         # For each frame, create a visual depth image from the inferenced depths.
         d_min, d_max = depths.min(), depths.max()
@@ -45,11 +46,14 @@ def stitch_rgbd_videos(
             rgb_full = full_frames[i]  # Full-resolution RGB frame.
             depth_frame = depths[i].astype(np.uint8)  # Reduce memory footprint
 
-            if d_max == d_min:
-                d_max = d_min + 1
+            print(f"Depth range: min={d_min}, max={d_max}, diff={d_max-d_min}")
+            
+            # Add a small buffer to ensure range is never zero
+            d_min_adj = max(0, d_min - 10)
+            d_max_adj = min(255, d_max + 10)
             
             # Normalize the depth frame to the range [0, 255].
-            depth_norm = ((depth_frame - d_min) / (d_max - d_min) * 255).astype(np.uint8)
+            depth_norm = ((depth_frame - d_min_adj) / (d_max_adj - d_min_adj) * 255).astype(np.uint8)
 
             depth_norm = np.clip(depth_norm, 0, 255)