File size: 1,674 Bytes
88590fc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
import spaces
import argparse

import os
import cv2
import torch
from insightface.app import FaceAnalysis
from imageio_ffmpeg import get_ffmpeg_exe

@spaces.GPU
def main(args):
    app = FaceAnalysis(
        providers=['CUDAExecutionProvider'],
        provider_options=[{'device_id': args.gpu_id}],
        root=args.insightface_model_path,
    )
    app.prepare(ctx_id=0, det_size=(args.height, args.width))

    os.system(f'{get_ffmpeg_exe()} -i "{args.video_path}" -y -vn "{args.audio_save_path}"')

    kps_sequence = []
    video_capture = cv2.VideoCapture(args.video_path)
    frame_idx = 0
    while video_capture.isOpened():
        ret, frame = video_capture.read()
        if not ret:
            break
        faces = app.get(frame)
        assert len(faces) == 1, f'There are {len(faces)} faces in the {frame_idx}-th frame. Only one face is supported.'

        kps = faces[0].kps[:3]
        kps_sequence.append(kps)
        frame_idx += 1
    torch.save(kps_sequence, args.kps_sequence_save_path)


if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--video_path', type=str, default='')
    parser.add_argument('--kps_sequence_save_path', type=str, default='')
    parser.add_argument('--audio_save_path', type=str, default='')
    parser.add_argument('--device', type=str, default='cuda')
    parser.add_argument('--gpu_id', type=int, default=0)
    parser.add_argument('--insightface_model_path', type=str, default='./model_ckpts/insightface_models/')
    parser.add_argument('--height', type=int, default=512)
    parser.add_argument('--width', type=int, default=512)
    args = parser.parse_args()

    main(args)