caps/extract_features_v3.py at master · zebin-dm/caps · 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
import torch
from torch.utils.data import Dataset
import os
import numpy as np
import cv2
import skimage.io as io
import torchvision.transforms as transforms
import config
from tqdm import tqdm
from CAPS.caps_model_v3 import CAPSModel
from imgaug import augmenters as iaa


def ratio_preserving_resize(img, target_size):
    '''
    :param img: raw img
    :param target_size: (h,w)
    :return:
    '''
    scales = np.array((target_size[0]/img.shape[0], target_size[1]/img.shape[1])) ##h_s,w_s

    new_size = np.round(np.array(img.shape)*np.max(scales)).astype(int)#
    temp_img = cv2.resize(img, tuple(new_size[::-1]))
    curr_h, curr_w = temp_img.shape
    target_h, target_w = target_size
    ##
    hp = (target_h-curr_h)//2
    wp = (target_w-curr_w)//2
    aug = iaa.Sequential([iaa.CropAndPad(px=(hp, wp, target_h-curr_h-hp, target_w-curr_w-wp),keep_size=False),])
    new_img = aug(images=temp_img)
    return new_img


def least_common_multiple(src_len, base_len):
    if src_len % base_len != 0:
        dst_len = (src_len // base_len + 1) * base_len
    else:
        dst_len = src_len
    return dst_len

def pic_rb_pading(img, base_len=16):
    sh, sw = img.shape
    dh = least_common_multiple(sh, base_len)
    dw = least_common_multiple(sw, base_len)

    if dh == sh and dw == sw:
        return img

    d_img = np.zeros((dh, dw), dtype=img.dtype)
    d_img[:sh, :sw] = img
    return d_img


class HPatchDataset(Dataset):
    def __init__(self, imdir):
        # self.transform = transforms.Compose([transforms.ToTensor(),
        #                                      transforms.Normalize(mean=(0.485, 0.456, 0.406),
        #                                                           std=(0.229, 0.224, 0.225)),
        #                                      ])
        self.transform = transforms.Compose([transforms.ToTensor()])
        self.imfs = []
        for f in os.listdir(imdir):
            scene_dir = os.path.join(imdir, f)
            self.imfs.extend([os.path.join(scene_dir, '{}.ppm').format(ind) for ind in range(1, 7)])

    def __getitem__(self, item):
        imf = self.imfs[item]
        # modify the loader from io.imread to cv2.imread
        # im = io.imread(imf)
        im = cv2.imread(imf, 0)
        src_shape = im.shape
        im = pic_rb_pading(im)
        # print(imf)
        # print("src_img")
        # print(im[100: 110, 100: 110])

        im_tensor = self.transform(im)
        # print("tesnor img shape: {}".format(im_tensor.shape))
        # print(im_tensor[0, 100: 110, 100: 110])
        # using sift keypoints
        # sift = cv2.SIFT_create()
        # gray = cv2.cvtColor(im, cv2.COLOR_RGB2GRAY)
        # kpts = sift.detect(im)
        # kpts = np.array([[kp.pt[0], kp.pt[1]] for kp in kpts])
        # coord = torch.from_numpy(kpts).float()
        # print("ori coord shape: {}".format(coord.shape))
        # out = {'im': im_tensor, 'coord': coord, 'imf': imf}
        # return out
        return im_tensor, src_shape, imf

    def __len__(self):
        return len(self.imfs)


if __name__ == '__main__':
    # example code for extracting features for HPatches dataset, SIFT keypoint is used
    args = config.get_args()
    device = torch.device('cuda:0')

    dataset = HPatchDataset(args.extract_img_dir)
    data_loader = torch.utils.data.DataLoader(dataset, batch_size=1, shuffle=False, num_workers=args.workers)

    model = CAPSModel(args)

    outdir = args.extract_out_dir
    os.makedirs(outdir, exist_ok=True)

    img_save_path = "/home/dm/work/02.workspace/caps/out/output_img"
    number = 0
    with torch.no_grad():
        for (im_data, src_shape, img_path) in data_loader:
            im_data = im_data.to(device)
            coord, feats = model.extract_det_and_des(im_data, src_shape)
            desc = feats[0].squeeze(0).detach().cpu().numpy()
            kpt = coord[0].cpu().numpy()

            save_folder = os.path.join(outdir, os.path.basename(os.path.dirname(img_path[0])))
            os.makedirs(save_folder, exist_ok=True)
            save_file = os.path.join(save_folder, "{}.magicv3_2".format(os.path.basename(img_path[0])))
            print(kpt.shape)
            print(desc.shape)
            break
            with open(save_file, 'wb') as output_file:
                np.savez(
                    output_file,
                    keypoints=kpt,
                    scores=[],
                    descriptors=desc)