File size: 4,549 Bytes
fca4fc0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
import os
root_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "..")
import sys
sys.path.append(root_dir)

import clip
import re
import argparse
import torch
import json
import numpy as np
from tqdm import tqdm
from torchvision.transforms import Compose, Resize, CenterCrop, Normalize
from vtimellm.model.builder import load_pretrained_model
from vtimellm.utils import disable_torch_init
from vtimellm.mm_utils import VideoExtractor
from glob import glob
import random

try:
    from torchvision.transforms import InterpolationMode
    BICUBIC = InterpolationMode.BICUBIC
except ImportError:
    from PIL import Image
    BICUBIC = Image.BICUBIC


def parse_args():
    parser = argparse.ArgumentParser()
    parser.add_argument("--clip_path", type=str, default="checkpoints/vtimellm/ViT-L-14.pt")
    parser.add_argument("--train_path", type=str, default="vtimellm/eval/data_example.json")
    parser.add_argument("--test_path", type=str, default="vtimellm/eval/data_example.json")
    parser.add_argument("--save_path", type=str, default="vtimellm/eval/data_example.json")

    parser.add_argument("--feat_folder", type=str, default=None)
    parser.add_argument("--video_folder", type=str, default=None)

    parser.add_argument("--merge", action='store_true')
    parser.add_argument("--merge_filename",type=str, default="vtimellm/eval/clipvitl14-vtimellm.pth")

    args = parser.parse_args()
    return args

if __name__ == "__main__":
    args = parse_args()
    disable_torch_init()

    save_path = args.save_path
    assert os.path.exists(save_path)

    if not args.merge:

        if args.video_folder is not None:

            print("Loading model..")
            clip_model, _ = clip.load(args.clip_path)
            clip_model.eval()
            clip_model = clip_model.cuda()
            print("Model load complete.")

            video_loader = VideoExtractor(N=100) # 100 frames

            transform = Compose([
                Resize(224, interpolation=BICUBIC),
                CenterCrop(224),
                Normalize((0.48145466, 0.4578275, 0.40821073), (0.26862954, 0.26130258, 0.27577711)),
            ])

        else:
            print("Provide me the video folder")
            assert False

        train = json.load(open(args.train_path))
        test = json.load(open(args.test_path))

        data_keys = list(train.keys()) + list(test.keys())
        data_keys = list(set(data_keys))
        random.shuffle(data_keys) # shuffle keys for each gpu to process different videos
        curr_saved = glob(f'{save_path}*.pth')

        print("*"*95)
        print(f'Save path: {save_path}')
        print(f'Num videos to extract: {len(data_keys)}')
        print(f'Currently saved features: {len(curr_saved)}')
        print("*"*95)

        for id in tqdm(data_keys):

            curr_saved = glob(f'{save_path}*.pth')
            curr_saved = [i.split('/')[-1][:-4] for i in curr_saved]

            if id not in curr_saved:
                features = None

                if features is None and args.video_folder is not None:
                    for ext in ['mp4', 'mkv', 'webm']:
                        video_path = os.path.join(args.video_folder, f"{id}.{ext}")
                        if os.path.isfile(video_path):
                            _, images = video_loader.extract({'id': None, 'video': video_path})
                            try:
                                images = transform(images / 255.0)
                                images = images.to(torch.float16)
                            except:
                                continue
                            with torch.no_grad():
                                features = clip_model.encode_image(images.to('cuda'))
                                break

                if features is None:
                    print(f"Failed to extract: {id}")
                    break
                    
                else:
                    torch.save(features.cpu(), f'{save_path}{id}.pth')
            else:
                print(f"Already exists {id}")
        print("Completed Extraction")


    else:
        video_features = {} # { video id : video feature }
        curr_saved = glob(f'{save_path}*.pth')

        for curr_path in curr_saved:
            vid_feature = torch.load(curr_path)

            # Get video id
            v_id = curr_path.split('/')[-1].split('.')[0]

            video_features[v_id] = vid_feature

        # save video feature
        torch.save(video_features, args.merge_filename)