Hugging Face's logo

Spaces:
YinuoGuo27
/
test_kdtalker
Running on Zero

App Files Community
YinuoGuo27 commited on
Commit
0560409
·
verified ·
1 Parent(s): a68c679

Upload 5 files

Browse files
Files changed (5) hide show
  1. difpoint/configs/onnx_infer.yaml +105 -0
  2. difpoint/configs/onnx_mp_infer.yaml +99 -0
  3. difpoint/configs/trt_infer.yaml +105 -0
  4. difpoint/configs/trt_mp_infer.yaml +99 -0
  5. difpoint/datasets/norm_info_d6.5_c8.5_vox1_train.npz +3 -0
difpoint/configs/onnx_infer.yaml ADDED
@@ -0,0 +1,105 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ models:
2
+ warping_spade:
3
+ name: "WarpingSpadeModel"
4
+ predict_type: "ort"
5
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/warping_spade.onnx"
6
+ motion_extractor:
7
+ name: "MotionExtractorModel"
8
+ predict_type: "ort"
9
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/motion_extractor.onnx"
10
+ landmark:
11
+ name: "LandmarkModel"
12
+ predict_type: "ort"
13
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/landmark.onnx"
14
+ face_analysis:
15
+ name: "FaceAnalysisModel"
16
+ predict_type: "ort"
17
+ model_path:
18
+ - "./difpoint/checkpoints/liveportrait_onnx/retinaface_det_static.onnx"
19
+ - "./difpoint/checkpoints/liveportrait_onnx/face_2dpose_106_static.onnx"
20
+ app_feat_extractor:
21
+ name: "AppearanceFeatureExtractorModel"
22
+ predict_type: "ort"
23
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/appearance_feature_extractor.onnx"
24
+ stitching:
25
+ name: "StitchingModel"
26
+ predict_type: "ort"
27
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/stitching.onnx"
28
+ stitching_eye_retarget:
29
+ name: "StitchingModel"
30
+ predict_type: "ort"
31
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/stitching_eye.onnx"
32
+ stitching_lip_retarget:
33
+ name: "StitchingModel"
34
+ predict_type: "ort"
35
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/stitching_lip.onnx"
36
+
37
+ animal_models:
38
+ warping_spade:
39
+ name: "WarpingSpadeModel"
40
+ predict_type: "ort"
41
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/warping_spade.onnx"
42
+ motion_extractor:
43
+ name: "MotionExtractorModel"
44
+ predict_type: "ort"
45
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/motion_extractor.onnx"
46
+ app_feat_extractor:
47
+ name: "AppearanceFeatureExtractorModel"
48
+ predict_type: "ort"
49
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/appearance_feature_extractor.onnx"
50
+ stitching:
51
+ name: "StitchingModel"
52
+ predict_type: "ort"
53
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/stitching.onnx"
54
+ stitching_eye_retarget:
55
+ name: "StitchingModel"
56
+ predict_type: "ort"
57
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/stitching_eye.onnx"
58
+ stitching_lip_retarget:
59
+ name: "StitchingModel"
60
+ predict_type: "ort"
61
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/stitching_lip.onnx"
62
+ landmark:
63
+ name: "LandmarkModel"
64
+ predict_type: "ort"
65
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/landmark.onnx"
66
+ face_analysis:
67
+ name: "FaceAnalysisModel"
68
+ predict_type: "ort"
69
+ model_path:
70
+ - "./difpoint/checkpoints/liveportrait_onnx/retinaface_det_static.onnx"
71
+ - "./difpoint/checkpoints/liveportrait_onnx/face_2dpose_106_static.onnx"
72
+
73
+ crop_params:
74
+ src_dsize: 512
75
+ src_scale: 2.3
76
+ src_vx_ratio: 0.0
77
+ src_vy_ratio: -0.125
78
+ dri_scale: 2.2
79
+ dri_vx_ratio: 0.0
80
+ dri_vy_ratio: -0.1
81
+
82
+
83
+ infer_params:
84
+ flag_crop_driving_video: False
85
+ flag_normalize_lip: True
86
+ flag_source_video_eye_retargeting: False
87
+ flag_video_editing_head_rotation: False
88
+ flag_eye_retargeting: False
89
+ flag_lip_retargeting: False
90
+ flag_stitching: True
91
+ flag_relative_motion: True
92
+ flag_pasteback: True
93
+ flag_do_crop: True
94
+ flag_do_rot: True
95
+
96
+ # NOT EXPOERTED PARAMS
97
+ lip_normalize_threshold: 0.03 # threshold for flag_normalize_lip
98
+ source_video_eye_retargeting_threshold: 0.18 # threshold for eyes retargeting if the input is a source video
99
+ driving_smooth_observation_variance: 1e-7 # smooth strength scalar for the animated video when the input is a source video, the larger the number, the smoother the animated video; too much smoothness would result in loss of motion accuracy
100
+ anchor_frame: 0 # TO IMPLEMENT
101
+ mask_crop_path: "./assets/mask_template.png"
102
+ driving_multiplier: 1.0
103
+
104
+ source_max_dim: 1280 # the max dim of height and width of source image
105
+ source_division: 2 # make sure the height and width of source image can be divided by this number
difpoint/configs/onnx_mp_infer.yaml ADDED
@@ -0,0 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ models:
2
+ warping_spade:
3
+ name: "WarpingSpadeModel"
4
+ predict_type: "ort"
5
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/warping_spade.onnx"
6
+ motion_extractor:
7
+ name: "MotionExtractorModel"
8
+ predict_type: "ort"
9
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/motion_extractor.onnx"
10
+ landmark:
11
+ name: "LandmarkModel"
12
+ predict_type: "ort"
13
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/landmark.onnx"
14
+ face_analysis:
15
+ name: "MediaPipeFaceModel"
16
+ predict_type: "mp"
17
+ app_feat_extractor:
18
+ name: "AppearanceFeatureExtractorModel"
19
+ predict_type: "ort"
20
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/appearance_feature_extractor.onnx"
21
+ stitching:
22
+ name: "StitchingModel"
23
+ predict_type: "ort"
24
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/stitching.onnx"
25
+ stitching_eye_retarget:
26
+ name: "StitchingModel"
27
+ predict_type: "ort"
28
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/stitching_eye.onnx"
29
+ stitching_lip_retarget:
30
+ name: "StitchingModel"
31
+ predict_type: "ort"
32
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/stitching_lip.onnx"
33
+
34
+ animal_models:
35
+ warping_spade:
36
+ name: "WarpingSpadeModel"
37
+ predict_type: "ort"
38
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/warping_spade.onnx"
39
+ motion_extractor:
40
+ name: "MotionExtractorModel"
41
+ predict_type: "ort"
42
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/motion_extractor.onnx"
43
+ app_feat_extractor:
44
+ name: "AppearanceFeatureExtractorModel"
45
+ predict_type: "ort"
46
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/appearance_feature_extractor.onnx"
47
+ stitching:
48
+ name: "StitchingModel"
49
+ predict_type: "ort"
50
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/stitching.onnx"
51
+ stitching_eye_retarget:
52
+ name: "StitchingModel"
53
+ predict_type: "ort"
54
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/stitching_eye.onnx"
55
+ stitching_lip_retarget:
56
+ name: "StitchingModel"
57
+ predict_type: "ort"
58
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/stitching_lip.onnx"
59
+ landmark:
60
+ name: "LandmarkModel"
61
+ predict_type: "ort"
62
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/landmark.onnx"
63
+ face_analysis:
64
+ name: "MediaPipeFaceModel"
65
+ predict_type: "mp"
66
+
67
+ crop_params:
68
+ src_dsize: 512
69
+ src_scale: 2.3
70
+ src_vx_ratio: 0.0
71
+ src_vy_ratio: -0.125
72
+ dri_scale: 2.2
73
+ dri_vx_ratio: 0.0
74
+ dri_vy_ratio: -0.1
75
+
76
+
77
+ infer_params:
78
+ flag_crop_driving_video: False
79
+ flag_normalize_lip: True
80
+ flag_source_video_eye_retargeting: False
81
+ flag_video_editing_head_rotation: False
82
+ flag_eye_retargeting: False
83
+ flag_lip_retargeting: False
84
+ flag_stitching: True
85
+ flag_relative_motion: True
86
+ flag_pasteback: True
87
+ flag_do_crop: True
88
+ flag_do_rot: True
89
+
90
+ # NOT EXPOERTED PARAMS
91
+ lip_normalize_threshold: 0.03 # threshold for flag_normalize_lip
92
+ source_video_eye_retargeting_threshold: 0.18 # threshold for eyes retargeting if the input is a source video
93
+ driving_smooth_observation_variance: 1e-7 # smooth strength scalar for the animated video when the input is a source video, the larger the number, the smoother the animated video; too much smoothness would result in loss of motion accuracy
94
+ anchor_frame: 0 # TO IMPLEMENT
95
+ mask_crop_path: "./assets/mask_template.png"
96
+ driving_multiplier: 1.0
97
+
98
+ source_max_dim: 1280 # the max dim of height and width of source image
99
+ source_division: 2 # make sure the height and width of source image can be divided by this number
difpoint/configs/trt_infer.yaml ADDED
@@ -0,0 +1,105 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ models:
2
+ warping_spade:
3
+ name: "WarpingSpadeModel"
4
+ predict_type: "trt"
5
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/warping_spade-fix.trt"
6
+ motion_extractor:
7
+ name: "MotionExtractorModel"
8
+ predict_type: "trt"
9
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/motion_extractor.trt"
10
+ landmark:
11
+ name: "LandmarkModel"
12
+ predict_type: "trt"
13
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/landmark.trt"
14
+ face_analysis:
15
+ name: "FaceAnalysisModel"
16
+ predict_type: "trt"
17
+ model_path:
18
+ - "./difpoint/checkpoints/liveportrait_onnx/retinaface_det_static.trt"
19
+ - "./difpoint/checkpoints/liveportrait_onnx/face_2dpose_106_static.trt"
20
+ app_feat_extractor:
21
+ name: "AppearanceFeatureExtractorModel"
22
+ predict_type: "trt"
23
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/appearance_feature_extractor.trt"
24
+ stitching:
25
+ name: "StitchingModel"
26
+ predict_type: "trt"
27
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/stitching.trt"
28
+ stitching_eye_retarget:
29
+ name: "StitchingModel"
30
+ predict_type: "trt"
31
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/stitching_eye.trt"
32
+ stitching_lip_retarget:
33
+ name: "StitchingModel"
34
+ predict_type: "trt"
35
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/stitching_lip.trt"
36
+
37
+ animal_models:
38
+ warping_spade:
39
+ name: "WarpingSpadeModel"
40
+ predict_type: "trt"
41
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/warping_spade-fix.trt"
42
+ motion_extractor:
43
+ name: "MotionExtractorModel"
44
+ predict_type: "trt"
45
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/motion_extractor.trt"
46
+ app_feat_extractor:
47
+ name: "AppearanceFeatureExtractorModel"
48
+ predict_type: "trt"
49
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/appearance_feature_extractor.trt"
50
+ stitching:
51
+ name: "StitchingModel"
52
+ predict_type: "trt"
53
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/stitching.trt"
54
+ stitching_eye_retarget:
55
+ name: "StitchingModel"
56
+ predict_type: "trt"
57
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/stitching_eye.trt"
58
+ stitching_lip_retarget:
59
+ name: "StitchingModel"
60
+ predict_type: "trt"
61
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/stitching_lip.trt"
62
+ landmark:
63
+ name: "LandmarkModel"
64
+ predict_type: "trt"
65
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/landmark.trt"
66
+ face_analysis:
67
+ name: "FaceAnalysisModel"
68
+ predict_type: "trt"
69
+ model_path:
70
+ - "./difpoint/checkpoints/liveportrait_onnx/retinaface_det_static.trt"
71
+ - "./difpoint/checkpoints/liveportrait_onnx/face_2dpose_106_static.trt"
72
+
73
+ crop_params:
74
+ src_dsize: 512
75
+ src_scale: 2.3
76
+ src_vx_ratio: 0.0
77
+ src_vy_ratio: -0.125
78
+ dri_scale: 2.2
79
+ dri_vx_ratio: 0.0
80
+ dri_vy_ratio: -0.1
81
+
82
+
83
+ infer_params:
84
+ flag_crop_driving_video: False
85
+ flag_normalize_lip: False
86
+ flag_source_video_eye_retargeting: False
87
+ flag_video_editing_head_rotation: False
88
+ flag_eye_retargeting: False
89
+ flag_lip_retargeting: False
90
+ flag_stitching: True
91
+ flag_relative_motion: True
92
+ flag_pasteback: True
93
+ flag_do_crop: True
94
+ flag_do_rot: True
95
+
96
+ # NOT EXPOERTED PARAMS
97
+ lip_normalize_threshold: 0.03 # threshold for flag_normalize_lip
98
+ source_video_eye_retargeting_threshold: 0.18 # threshold for eyes retargeting if the input is a source video
99
+ driving_smooth_observation_variance: 1e-7 # smooth strength scalar for the animated video when the input is a source video, the larger the number, the smoother the animated video; too much smoothness would result in loss of motion accuracy
100
+ anchor_frame: 0 # TO IMPLEMENT
101
+ mask_crop_path: "./assets/mask_template.png"
102
+ driving_multiplier: 1.0
103
+
104
+ source_max_dim: 1280 # the max dim of height and width of source image
105
+ source_division: 2 # make sure the height and width of source image can be divided by this number
difpoint/configs/trt_mp_infer.yaml ADDED
@@ -0,0 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ models:
2
+ warping_spade:
3
+ name: "WarpingSpadeModel"
4
+ predict_type: "trt"
5
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/warping_spade-fix.trt"
6
+ motion_extractor:
7
+ name: "MotionExtractorModel"
8
+ predict_type: "trt"
9
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/motion_extractor.trt"
10
+ landmark:
11
+ name: "LandmarkModel"
12
+ predict_type: "trt"
13
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/landmark.trt"
14
+ face_analysis:
15
+ name: "MediaPipeFaceModel"
16
+ predict_type: "mp"
17
+ app_feat_extractor:
18
+ name: "AppearanceFeatureExtractorModel"
19
+ predict_type: "trt"
20
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/appearance_feature_extractor.trt"
21
+ stitching:
22
+ name: "StitchingModel"
23
+ predict_type: "trt"
24
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/stitching.trt"
25
+ stitching_eye_retarget:
26
+ name: "StitchingModel"
27
+ predict_type: "trt"
28
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/stitching_eye.trt"
29
+ stitching_lip_retarget:
30
+ name: "StitchingModel"
31
+ predict_type: "trt"
32
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/stitching_lip.trt"
33
+
34
+ animal_models:
35
+ warping_spade:
36
+ name: "WarpingSpadeModel"
37
+ predict_type: "trt"
38
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/warping_spade-fix.trt"
39
+ motion_extractor:
40
+ name: "MotionExtractorModel"
41
+ predict_type: "trt"
42
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/motion_extractor.trt"
43
+ app_feat_extractor:
44
+ name: "AppearanceFeatureExtractorModel"
45
+ predict_type: "trt"
46
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/appearance_feature_extractor.trt"
47
+ stitching:
48
+ name: "StitchingModel"
49
+ predict_type: "trt"
50
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/stitching.trt"
51
+ stitching_eye_retarget:
52
+ name: "StitchingModel"
53
+ predict_type: "trt"
54
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/stitching_eye.trt"
55
+ stitching_lip_retarget:
56
+ name: "StitchingModel"
57
+ predict_type: "trt"
58
+ model_path: "./difpoint/checkpoints/liveportrait_animal_onnx/stitching_lip.trt"
59
+ landmark:
60
+ name: "LandmarkModel"
61
+ predict_type: "trt"
62
+ model_path: "./difpoint/checkpoints/liveportrait_onnx/landmark.trt"
63
+ face_analysis:
64
+ name: "MediaPipeFaceModel"
65
+ predict_type: "mp"
66
+
67
+ crop_params:
68
+ src_dsize: 512
69
+ src_scale: 2.0
70
+ src_vx_ratio: 0.0
71
+ src_vy_ratio: -0.125
72
+ dri_scale: 2.2
73
+ dri_vx_ratio: 0.0
74
+ dri_vy_ratio: -0.1
75
+
76
+
77
+ infer_params:
78
+ flag_crop_driving_video: False
79
+ flag_normalize_lip: False
80
+ flag_source_video_eye_retargeting: False
81
+ flag_video_editing_head_rotation: False
82
+ flag_eye_retargeting: False
83
+ flag_lip_retargeting: False
84
+ flag_stitching: False
85
+ flag_relative_motion: False
86
+ flag_pasteback: False
87
+ flag_do_crop: False
88
+ flag_do_rot: False
89
+
90
+ # NOT EXPOERTED PARAMS
91
+ lip_normalize_threshold: 0.03 # threshold for flag_normalize_lip
92
+ source_video_eye_retargeting_threshold: 0.18 # threshold for eyes retargeting if the input is a source video
93
+ driving_smooth_observation_variance: 1e-7 # smooth strength scalar for the animated video when the input is a source video, the larger the number, the smoother the animated video; too much smoothness would result in loss of motion accuracy
94
+ anchor_frame: 0 # TO IMPLEMENT
95
+ mask_crop_path: "./assets/mask_template.png"
96
+ driving_multiplier: 1.0
97
+
98
+ source_max_dim: 1280 # the max dim of height and width of source image
99
+ source_division: 2 # make sure the height and width of source image can be divided by this number
difpoint/datasets/norm_info_d6.5_c8.5_vox1_train.npz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9422e503e75df9d1bd455d8e0f9f5e2826b12956cdedbb5566097c0151bddafb
3
+ size 5580