-
Notifications
You must be signed in to change notification settings - Fork 12
/
visualize_multilabel.py
270 lines (238 loc) · 11.4 KB
/
visualize_multilabel.py
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
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
import os
import sys
import argparse
import numpy as np
import cv2
import PIL
from PIL import Image
import matplotlib.pyplot as plt
import zipfile
import shutil
import h5py
import torch
from torch import sigmoid
import torchvision.transforms as transforms
import torchvision.datasets as datasets
# import torch.backends.cudnn as cudnn
import torch.nn.functional as F
from modules.CASENet import CASENet_resnet101
from prep_dataset.prep_cityscapes_dataset import RGB2BGR, ToTorchFormatTensor
import utils.utils as utils
def get_cityscapes_class_names():
return ['road',
'sidewalk',
'building',
'wall',
'fence',
'pole',
'traffic light',
'traffic sign',
'vegetation',
'terrain',
'sky',
'person',
'rider',
'car',
'truck',
'bus',
'train',
'motorcycle',
'bicycle']
def normalized_feature_map(fmap):
fmap_min = fmap.min()
fmap_max = fmap.max()
fmap = (fmap-fmap_min)/(fmap_max-fmap_min)
return fmap
# color map for each trainId, from the official cityscapes script
# https://github.com/mcordts/cityscapesScripts/blob/master/cityscapesscripts/helpers/labels.py
def get_colors():
color_dict = {}
color_dict[0] = [128, 64, 128]
color_dict[1] = [244, 35, 232]
color_dict[2] = [70, 70, 70]
color_dict[3] = [102, 102, 156]
color_dict[4] = [190, 153, 153]
color_dict[5] = [153, 153, 153]
color_dict[6] = [250, 170, 30]
color_dict[7] = [220, 220, 0]
color_dict[8] = [107, 142, 35]
color_dict[9] = [152, 251, 152]
color_dict[10] = [70, 130, 180]
color_dict[11] = [220, 20, 60]
color_dict[12] = [255, 0, 0]
color_dict[13] = [0, 0, 142]
color_dict[14] = [0, 0, 70]
color_dict[15] = [0, 60, 100]
color_dict[16] = [0, 80, 100]
color_dict[17] = [0, 0, 230]
color_dict[18] = [119, 11, 32]
return color_dict
if __name__ == "__main__":
parser = argparse.ArgumentParser(sys.argv[0])
parser.add_argument('-m', '--model', type=str,
help="path to the Pytorch model(.pth) containing the trained weights")
parser.add_argument('-l', '--image_list', type=str, default='',
help="list of image files to be tested")
parser.add_argument('-f', '--image_file', type=str, default='',
help="a single image file to be tested")
parser.add_argument('-d', '--image_dir', type=str, default='',
help="root folder of the image files in the list or the single image file")
parser.add_argument('-o', '--output_dir', type=str, default='.',
help="folder to store the test results")
args = parser.parse_args(sys.argv[1:])
# load input path
if os.path.exists(args.image_list):
with open(args.image_list) as f:
ori_test_list = [x.strip().split()[0] for x in f.readlines()]
if args.image_dir!='':
test_list = [
args.image_dir+x if os.path.isabs(x)
else os.path.join(args.image_dir, x)
for x in ori_test_list]
else:
image_file = os.path.join(args.image_dir, args.image_file)
if os.path.exists(image_file):
ori_test_list = [args.image_file]
test_list = [image_file]
else:
raise IOError('nothing to be tested!')
# load network
num_cls = 19
model = CASENet_resnet101(pretrained=False, num_classes=num_cls)
# model = model.cuda()
model = model.eval()
# cudnn.benchmark = True
utils.load_pretrained_model(model, args.model)
cls_names = get_cityscapes_class_names()
color_dict = get_colors()
if not os.path.exists(args.output_dir):
os.makedirs(args.output_dir)
# Define normalization for data
input_size = 472
normalize = transforms.Normalize(mean=[104.008, 116.669, 122.675], std=[1, 1, 1])
img_transform = transforms.Compose([
transforms.Resize([input_size, input_size]),
RGB2BGR(roll=True),
ToTorchFormatTensor(div=False),
normalize,
])
label_transform = transforms.Compose([
transforms.ToPILImage(),
transforms.Resize([input_size, input_size], interpolation=PIL.Image.NEAREST),
transforms.ToTensor(),
])
h5_f = h5py.File("./utils/val_label_binary_np.h5", 'r')
for idx_img in range(len(test_list)):
img = Image.open(test_list[idx_img]).convert('RGB')
processed_img = img_transform(img).unsqueeze(0)
processed_img = utils.check_gpu(None, processed_img)
score_feats1, score_feats2, score_feats3, score_feats5, score_fuse_feats = model(processed_img, for_vis=True)
# Load numpy from hdf5 for gt.
np_data = h5_f['data/'+ori_test_list[idx_img].replace('leftImg8bit', 'gtFine').replace('/', '_').replace('.png', '_edge.npy')]
label_data = []
num_cls = np_data.shape[2]
for k in range(num_cls):
if np_data[:,:,num_cls-1-k].sum() > 0:
label_tensor = label_transform(torch.from_numpy(np_data[:, :, num_cls-1-k]).unsqueeze(0).float())
else: # ALL zeros, don't need transform, maybe a bit faster?..
label_tensor = torch.zeros(1, input_size, input_size).float()
label_data.append(label_tensor.squeeze(0).long())
label_data = torch.stack(label_data).transpose(0,1).transpose(1,2) # N X H X W -> H X W X N
img_base_name_noext = os.path.splitext(os.path.basename(test_list[idx_img]))[0]
img_base_name_noext = img_base_name_noext.replace('_leftImg8bit', '')
score_feats_list = [score_feats1, score_feats2, score_feats3]
score_feats_str_list = ['feats1', 'feats2', 'feats3']
# visualize side edge activation
for i in range(len(score_feats_list)):
feature = score_feats_list[i]
feature_str = score_feats_str_list[i]
side = normalized_feature_map(feature.data[0][0, :, :].cpu().numpy())
im = (side*255).astype(np.uint8)
if not os.path.exists(os.path.join(args.output_dir, img_base_name_noext)):
os.makedirs(os.path.join(args.output_dir, img_base_name_noext))
cv2.imwrite(
os.path.join(args.output_dir, img_base_name_noext, img_base_name_noext+'_'+feature_str+'.png'),
im)
# visualize side class activation
side_cls = normalized_feature_map(np.transpose(score_feats5.data[0].cpu().numpy(), (1, 2, 0)))
for idx_cls in range(num_cls):
side_cls_i = side_cls[:, :, idx_cls]
im = (side_cls_i * 255).astype(np.uint8)
if not os.path.exists(os.path.join(args.output_dir, img_base_name_noext)):
os.makedirs(os.path.join(args.output_dir, img_base_name_noext))
cv2.imwrite(
os.path.join(args.output_dir, img_base_name_noext, img_base_name_noext+'_'+'feats5'+'_'+cls_names[num_cls-idx_cls-1]+'.png'),
im)
# visualize predicted class and contours
score_output = sigmoid(score_fuse_feats.transpose(1,3).transpose(1,2)).data[0].cpu().numpy()
for idx_cls in range(num_cls):
r = np.zeros((score_output.shape[0], score_output.shape[1]))
g = np.zeros((score_output.shape[0], score_output.shape[1]))
b = np.zeros((score_output.shape[0], score_output.shape[1]))
rgb = np.zeros((score_output.shape[0], score_output.shape[1], 3))
score_pred = score_output[:, :, idx_cls]
score_pred_flag = (score_pred>0.5)
r[score_pred_flag==1] = color_dict[idx_cls][0]
g[score_pred_flag==1] = color_dict[idx_cls][1]
b[score_pred_flag==1] = color_dict[idx_cls][2]
r[score_pred_flag==0] = 255
g[score_pred_flag==0] = 255
b[score_pred_flag==0] = 255
rgb[:,:,0] = (r/255.0)
rgb[:,:,1] = (g/255.0)
rgb[:,:,2] = (b/255.0)
if not os.path.exists(os.path.join(args.output_dir, img_base_name_noext)):
os.makedirs(os.path.join(args.output_dir, img_base_name_noext))
plt.imsave(os.path.join(args.output_dir, img_base_name_noext, img_base_name_noext+'_fused_pred_'+cls_names[idx_cls]+'.png'), rgb)
gray = np.zeros(r.shape, dtype=np.uint8)
gray[score_pred_flag==0] = 0
gray[score_pred_flag==1] = 127
# create border for image
top = 1
bottom = top
left = 1
right = left
gray = cv2.copyMakeBorder(gray, top, bottom, left, right, cv2.BORDER_CONSTANT, None, (127,))
# morphological operator
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(7,7))
# dilated = cv2.dilate(gray, kernel)
closed = cv2.morphologyEx(gray, cv2.MORPH_CLOSE, kernel) # Closing Morphological Operator
# opened = cv2.morphologyEx(gray, cv2.MORPH_OPEN, kernel) # Opening Morphological Operator
_, cnts, hierarchy = cv2.findContours(closed.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
# select contour based on area and arc length
# print(len(cnts))
# cnts = [c for c in cnts if (cv2.arcLength(c, True) < 1880.0 and cv2.arcLength(c, True) >= 10.0)]
cnts = [c for c in cnts if (cv2.arcLength(c, True) < 1880.0 and cv2.contourArea(c, True) >= 150.0)]
cnts = sorted(cnts, key = cv2.contourArea, reverse = True)
# print(cls_names[idx_cls])
# for c in cnts:
# print(cv2.arcLength(c, True), cv2.contourArea(c))
# cv2.drawContours(gray, cnts, -1, (255,255,0), 3)
resizedimg = img.resize((472,472), resample=PIL.Image.BILINEAR)
opencv_image = np.array(resizedimg)
opencv_image = opencv_image[:, :, ::-1].copy()
cv2.drawContours(opencv_image, cnts, -1, (255,255,0), 1)
opencv_image = cv2.resize(opencv_image, img.size, interpolation = cv2.INTER_CUBIC)
if not os.path.exists(os.path.join(args.output_dir, img_base_name_noext)):
os.makedirs(os.path.join(args.output_dir, img_base_name_noext))
cv2.imwrite(os.path.join(args.output_dir, img_base_name_noext, img_base_name_noext+'_'+cls_names[idx_cls]+'.png'), opencv_image)
# gt visualization
gt_data = label_data.numpy()
for idx_cls in range(num_cls):
r = np.zeros((gt_data.shape[0], gt_data.shape[1]))
g = np.zeros((gt_data.shape[0], gt_data.shape[1]))
b = np.zeros((gt_data.shape[0], gt_data.shape[1]))
rgb = np.zeros((gt_data.shape[0], gt_data.shape[1], 3))
score_pred_flag = gt_data[:, :, idx_cls]
r[score_pred_flag==1] = color_dict[idx_cls][0]
g[score_pred_flag==1] = color_dict[idx_cls][1]
b[score_pred_flag==1] = color_dict[idx_cls][2]
r[score_pred_flag==0] = 255
g[score_pred_flag==0] = 255
b[score_pred_flag==0] = 255
rgb[:,:,0] = (r/255.0)
rgb[:,:,1] = (g/255.0)
rgb[:,:,2] = (b/255.0)
plt.imsave(os.path.join(args.output_dir, img_base_name_noext, img_base_name_noext+'_gt_'+cls_names[idx_cls]+'.png'), rgb)
print('processed: '+test_list[idx_img])
print('Done!')