segmentation_benchmark/dataset_ood.py at main · rainarit/segmentation_benchmark · GitHub

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
import datetime
import os
import time
import csv
import shutil
import numpy as np
from tqdm import tqdm

import torch
import torch.utils.data
import torchvision
import semseg.transforms as T
import semseg.utils as utils
from torchvision.utils import save_image
from semseg.coco_utils import get_coco

class SegmentationPresetEval:
    def __init__(self, base_size, mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225), brightness=1, contrast=1, saturation=1, kernel_size=1, sigma=(0.1, 2.0)):
        self.transforms = T.Compose(
            [
                T.RandomResize(base_size, base_size),
                T.PILToTensor(),
                T.ConvertImageDtype(torch.float),
                T.Normalize(mean=mean, std=std),
                T.ColorJitter(brightness=brightness, contrast=contrast, saturation=saturation),
                #T.GaussanBlur(kernel_size, sigma),
            ]
        )

    def __call__(self, img, target):
        return self.transforms(img, target)

def get_dataset(dir_path, name, image_set, transform):
    def sbd(*args, **kwargs):
        return torchvision.datasets.SBDataset(*args, mode="segmentation", **kwargs)

    paths = {
        "voc": (dir_path, torchvision.datasets.VOCSegmentation, 21),
        "voc_aug": (dir_path, sbd, 21),
        "coco": (dir_path, get_coco, 21),
    }
    p, ds_fn, num_classes = paths[name]

    ds = ds_fn(p, image_set=image_set, transforms=transform)
    return ds, num_classes

def get_transform(brightness=0, contrast=0, saturation=0, kernel_size=0, sigma=(0.1, 2.0)):
    return SegmentationPresetEval(base_size=520, mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225), brightness=brightness, contrast=contrast, saturation=saturation, kernel_size=kernel_size, sigma=sigma)

def main(args):
    utils.init_distributed_mode(args)
    print(args)

    os.makedirs(args.root)

    if args.dataset == 'voc_aug':
        val_txt = os.path.join(args.root, 'val.txt')
        shutil.copy2('benchmark_RELEASE/dataset/val.txt', val_txt)

    print("Starting OOD dataset creation\n")
    print("="*70)

    # contrast
    print("\nOOD: Contrast\n")
    print("="*70)
    args.contrast_dir = os.path.join(args.root, "contrast")
    os.makedirs(args.contrast_dir)
    for contrast in range(1,6):
        print(f"\nContrast: {contrast}\n")
        print("="*35)
        args.contrast_severity_dir = os.path.join(args.contrast_dir, f'{contrast}')
        os.makedirs(args.contrast_severity_dir)
        dataset, _ = get_dataset(args.data_path, args.dataset, "val", get_transform(contrast=contrast))
        data_loader = torch.utils.data.DataLoader(dataset, batch_size=1, num_workers=args.workers, collate_fn=utils.collate_fn)
        for i, (image, target) in tqdm(enumerate(data_loader)):
            f = open(val_txt)
            content = f.readlines()
            save_image(image[0], str(os.path.join(args.contrast_severity_dir, f'{content[i]}.png')))

    # brightness
    print("\nOOD: Brightness\n")
    print("="*70)
    args.brightness_dir = os.path.join(args.root, "brightness")
    os.makedirs(args.brightness_dir)
    for brightness in range(1,6):
        print(f"\nBrightness: {brightness}\n")
        print("="*35)
        args.brightness_severity_dir = os.path.join(args.brightness_dir, f'{brightness}')
        os.makedirs(args.brightness_severity_dir)
        dataset, _ = get_dataset(args.data_path, args.dataset, "val", get_transform(brightness=brightness))
        data_loader = torch.utils.data.DataLoader(dataset, batch_size=1, num_workers=args.workers, collate_fn=utils.collate_fn)
        for i, (image, target) in enumerate(data_loader):
            f = open(val_txt)
            content = f.readlines()
            save_image(image[0], str(os.path.join(args.brightness_severity_dir, f'{content[i]}.png')))

    # gaussian blur
    # print("\nOOD: Gaussian Blur\n")
    # print("="*70)
    # args.gaussian_blur_dir = os.path.join(args.root, "gaussian_blur")
    # os.makedirs(args.gaussian_blur_dir)
    # for gaussian_blur in range(2,5):
    #     args.gaussian_blur_severity_dir = os.path.join(args.gaussian_blur_dir, f'{gaussian_blur}')
    #     os.makedirs(args.gaussian_blur_severity_dir)
    #     dataset, _ = get_dataset(args.data_path, args.dataset, "val", get_transform(gaussian_blur=gaussian_blur))
    #     data_loader = torch.utils.data.DataLoader(dataset, batch_size=1, num_workers=args.workers, collate_fn=utils.collate_fn)
    #     for i, (image, target) in enumerate(data_loader):
    #         f = open(val_txt)
    #         content = f.readlines()
    #         save_image(image[0], str(os.path.join(args.gaussian_blur_severity_dir, f'{content[i]}.png')))

def get_args_parser(add_help=True):
    import argparse

    parser = argparse.ArgumentParser(description="OOD Dataset Creation", add_help=add_help)

    parser.add_argument("--data-path", default="/home/AD/rraina/segmentation_benchmark/benchmark_RELEASE/dataset", type=str, help="dataset path")
    parser.add_argument("--dataset", default="voc_aug", type=str, help="dataset name")
    parser.add_argument("--root", default="SBD_OOD", type=str, help="ood dataset name")
    parser.add_argument(
        "-j", "--workers", default=16, type=int, metavar="N", help="number of data loading workers (default: 16)"
    )
    # distributed training parameters
    parser.add_argument("--world-size", default=1, type=int, help="number of distributed processes")
    parser.add_argument("--dist-url", default="env://", type=str, help="url used to set up distributed training")
    return parser


if __name__ == '__main__':
    args = get_args_parser().parse_args()
    main(args)