File size: 4,220 Bytes
94bafa8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from .sky_datasets import Sky
from torchvision import transforms
from .taichi_datasets import Taichi
from datasets import video_transforms
from .ucf101_datasets import UCF101
from .ffs_datasets import FaceForensics
from .ffs_image_datasets import FaceForensicsImages
from .sky_image_datasets import SkyImages
from .ucf101_image_datasets import UCF101Images
from .taichi_image_datasets import TaichiImages


def get_dataset(args):
    temporal_sample = video_transforms.TemporalRandomCrop(args.num_frames * args.frame_interval) # 16 1

    if args.dataset == 'ffs':
        transform_ffs = transforms.Compose([
            video_transforms.ToTensorVideo(), # TCHW
            video_transforms.RandomHorizontalFlipVideo(),
            video_transforms.UCFCenterCropVideo(args.image_size),
            transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True)
        ])
        return FaceForensics(args, transform=transform_ffs, temporal_sample=temporal_sample)
    elif args.dataset == 'ffs_img':
        transform_ffs = transforms.Compose([
            video_transforms.ToTensorVideo(), # TCHW
            video_transforms.RandomHorizontalFlipVideo(),
            video_transforms.UCFCenterCropVideo(args.image_size),
            transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True)
        ])
        return FaceForensicsImages(args, transform=transform_ffs, temporal_sample=temporal_sample)
    elif args.dataset == 'ucf101':
        transform_ucf101 = transforms.Compose([
            video_transforms.ToTensorVideo(), # TCHW
            video_transforms.RandomHorizontalFlipVideo(),
            video_transforms.UCFCenterCropVideo(args.image_size),
            transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True)
        ])
        return UCF101(args, transform=transform_ucf101, temporal_sample=temporal_sample)
    elif args.dataset == 'ucf101_img':
        transform_ucf101 = transforms.Compose([
            video_transforms.ToTensorVideo(), # TCHW
            video_transforms.RandomHorizontalFlipVideo(),
            video_transforms.UCFCenterCropVideo(args.image_size),
            transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True)
        ])
        return UCF101Images(args, transform=transform_ucf101, temporal_sample=temporal_sample)
    elif args.dataset == 'taichi':
        transform_taichi = transforms.Compose([
            video_transforms.ToTensorVideo(), # TCHW
            video_transforms.RandomHorizontalFlipVideo(),
            transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True)
        ])
        return Taichi(args, transform=transform_taichi, temporal_sample=temporal_sample)
    elif args.dataset == 'taichi_img':
        transform_taichi = transforms.Compose([
            video_transforms.ToTensorVideo(), # TCHW
            video_transforms.RandomHorizontalFlipVideo(),
            transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True)
        ])
        return TaichiImages(args, transform=transform_taichi, temporal_sample=temporal_sample)
    elif args.dataset == 'sky':  
        transform_sky = transforms.Compose([
                    video_transforms.ToTensorVideo(),
                    video_transforms.CenterCropResizeVideo(args.image_size),
                    # video_transforms.RandomHorizontalFlipVideo(),
                    transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True)
            ])
        return Sky(args, transform=transform_sky, temporal_sample=temporal_sample)
    elif args.dataset == 'sky_img':  
        transform_sky = transforms.Compose([
                    video_transforms.ToTensorVideo(),
                    video_transforms.CenterCropResizeVideo(args.image_size),
                    # video_transforms.RandomHorizontalFlipVideo(),
                    transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True)
            ])
        return SkyImages(args, transform=transform_sky, temporal_sample=temporal_sample)
    else:
        raise NotImplementedError(args.dataset)