CS-Net/predict3d.py at master · iMED-Lab/CS-Net · 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
132
133
134
135
136
137
138
139
import torch
import torch.nn.functional as F
import numpy as np
import os
import glob
from tqdm import tqdm
import SimpleITK as sitk
from utils.misc import get_spacing

os.environ['CUDA_VISIBLE_DEVICES'] = "1"

DATABASE = 'VascuSynth3/'
#
args = {
    'root'     : './dataset/' + DATABASE,
    'test_path': './dataset/' + DATABASE + 'test/',
    'pred_path': 'assets/' + 'VascuSynth3/',
    'img_size' : 512
}

if not os.path.exists(args['pred_path']):
    os.makedirs(args['pred_path'])


def rescale(img):
    w, h = img.size
    min_len = min(w, h)
    new_w, new_h = min_len, min_len
    scale_w = (w - new_w) // 2
    scale_h = (h - new_h) // 2
    box = (scale_w, scale_h, scale_w + new_w, scale_h + new_h)
    img = img.crop(box)
    return img


def load_3d():
    test_images = []
    test_labels = []
    for file in glob.glob(os.path.join(args['test_path'], 'images', '*.mha')):
        basename = os.path.basename(file)
        file_name = basename[:-8]
        image_name = os.path.join(args['test_path'], 'images', basename)
        label_name = os.path.join(args['test_path'], 'label', file_name + 'gt.mha')
        test_images.append(image_name)
        test_labels.append(label_name)
    return test_images, test_labels


def load_net():
    net = torch.load('/home/imed/Research/Attention/checkpoint/model.pkl')
    return net


def save_prediction(pred, filename='', spacing=None):
    pred = torch.argmax(pred, dim=1)
    save_path = args['pred_path'] + 'pred/'
    if not os.path.exists(save_path):
        os.makedirs(save_path)
        print("Make dirs success!")
    # for MSELoss()
    mask = (pred.data.cpu().numpy() * 255).astype(np.uint8)

    # thresholding
    # mask[mask >= 100] = 255
    # mask[mask < 100] = 0

    # mask = (mask.squeeze(0)).squeeze(0)  # 3D numpy array
    mask = mask.squeeze(0)  # for CE Loss
    # image = nib.Nifti1Image(np.int32(mask), affine)
    # nib.save(image, save_path + filename + ".nii.gz")
    mask = sitk.GetImageFromArray(mask)
    # if spacing is not None:
    #     mask.SetSpacing(spacing)
    sitk.WriteImage(mask, os.path.join(save_path + filename + ".mha"))


def save_probability(pred, label, filename=""):
    save_path = args['pred_path'] + 'pred/'
    if not os.path.exists(save_path):
        os.makedirs(save_path)
        print("Make dirs success!")
    # # for MSELoss()
    # mask = (pred.data.cpu().numpy() * 255)  # .astype(np.uint8)
    #
    # mask = mask.squeeze(0)
    # class0 = mask[0, :, :, :]
    # class1 = mask[1, :, :, :]
    # label = label / 255
    # class0 = class0 * label
    # class1 = class1 * label
    #
    # probability = class0 + class1

    probability = F.softmax(pred, dim=1)
    probability.squeeze_(0)
    class0 = probability[0, :, :, :]
    class1 = probability[1, :, :, :]
    class0 = sitk.GetImageFromArray(class0)
    class1 = sitk.GetImageFromArray(class1)
    sitk.WriteImage(class1, os.path.join(save_path + filename + "class1.mha"))


def save_label(label, index, spacing=None):
    label_path = args['pred_path'] + 'label/'
    if not os.path.exists(label_path):
        os.makedirs(label_path)
    label = sitk.GetImageFromArray(label)
    if spacing is not None:
        label.SetSpacing(spacing)
    sitk.WriteImage(label, os.path.join(label_path, index + ".mha"))


def predict():
    net = load_net()
    images, labels = load_3d()
    with torch.no_grad():
        net.eval()
        for i in tqdm(range(len(images))):
            name_list = images[i].split('/')
            index = name_list[-1][:-4]
            image = sitk.ReadImage(images[i])
            image = sitk.GetArrayFromImage(image).astype(np.float32)
            image = image / 255
            label = sitk.ReadImage(labels[i])
            label = sitk.GetArrayFromImage(label).astype(np.int64)
            # label = label / 255
            # VascuSynth
            # image = image[2:98, 2:98, 2:98]
            # label = label[2:98, 2:98, 2:98]
            save_label(label, index)
            # if cuda
            image = torch.from_numpy(np.ascontiguousarray(image)).unsqueeze(0).unsqueeze(0)
            image = image.cuda()
            output = net(image)
            save_prediction(output, filename=index + '_pred', spacing=None)


if __name__ == '__main__':
    predict()