-
Notifications
You must be signed in to change notification settings - Fork 0
/
visualize.py
223 lines (191 loc) · 8.59 KB
/
visualize.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
import os, glob
import json
import torch
import cv2
import numpy as np
import matplotlib.pyplot as plt
from PIL import Image
# from automatic_label_demo import show_box_tokens
def show_box_tokens(box,ax,token_pairs):
x0, y0 = box[0], box[1]
w, h = box[2] - box[0], box[3] - box[1]
ax.add_patch(plt.Rectangle((x0, y0), w, h, edgecolor='green', facecolor=(0,0,0,0), lw=2))
# ax.text(x0, y0, label)
labels = ''
for token, score in token_pairs.items():
labels += '{}({:.3f}) '.format(token, score)
ax.text(x0, y0, labels)
def show_mask(mask, ax, random_color=False):
if random_color:
color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)
else:
color = np.array([30/255, 144/255, 255/255, 0.6])
h, w = mask.shape[-2:]
mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
ax.imshow(mask_image)
def draw_object_detection(rgb:np.ndarray, box:np.ndarray, label, color=(0,255,0), thickness=2):
x0, y0 = box[0], box[1]
x1, y1 = box[2], box[3]
cv2.rectangle(rgb, (x0, y0), (x1, y1), color, thickness)
cv2.putText(rgb, label, (x0+5, y0 +16), cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 2)
return rgb
def read_scan_pairs(dir):
with open(dir) as f:
scan_pairs = []
for line in f.readlines():
scan_pairs.append(line.strip().split(' '))
f.close()
return scan_pairs
def read_scan_list(dir):
with open(dir) as f:
scan_list = f.readlines()
scans = [scan.strip() for scan in scan_list]
f.close()
return scans
def load_pred(label_file,valid_openset_names=None):
with open(label_file, 'r') as f:
json_data = json.load(f)
tags = json_data['tags'] if 'tags' in json_data else ''
raw_tags = json_data['raw_tags'] if 'raw_tags' in json_data else ''
masks = json_data['mask']
boxes = [] # [x1,y1,x2,y2]
semantics = [] # [{label:conf}]
for ele in masks:
if 'box' in ele:
# if label[-1]==',':label=label[:-1]
instance_id = ele['value']-1
detection_id = ele['value']
bbox = ele['box']
labels = ele['labels'] # {label:conf}
# box_area_normal = (bbox[2]-bbox[0])*(bbox[3]-bbox[1])/(img_width*img_height)
# if box_area_normal > MAX_BOX_RATIO: continue
if valid_openset_names is not None:
valid = False
for label in labels:
if label in valid_openset_names:
valid = True
break
if valid==False: continue
box_tensor = torch.Tensor([[bbox[0],bbox[1]],[bbox[2],bbox[3]]]) # [x1,y1,x2,y2]
boxes.append(box_tensor.unsqueeze(0)) # [x1,y1,x2,y2]
semantics.append(labels)
# z_ = Detection(bbox[0],bbox[1],bbox[2],bbox[3],labels)
# z_.add_mask(mask==detection_id)
# detections.append(z_)
else: # background
continue
assert len(boxes) == len(masks)-1, 'boxes dimension not aligned'
f.close()
if len(boxes)>0:
boxes = torch.cat(boxes, dim=0) # (num_prompts, 2, 2)
LINE_LENGTH = 60
if len(raw_tags)>LINE_LENGTH:
# seperate tags by line. Each line should be less than 50 characters
number_lines = int(len(raw_tags)/LINE_LENGTH)+1
rephrase_raw_tags = ''
for i in range(number_lines):
rephrase_raw_tags += raw_tags[i*LINE_LENGTH:(i+1)*LINE_LENGTH]+'\n'
else:
rephrase_raw_tags = raw_tags
joint_tags = 'raw tags: {} valid tags: {}'.format(rephrase_raw_tags, tags)
return boxes, semantics, joint_tags
else:
return torch.zeros((1,2,2)), [], ''
return None,None,None
if __name__=='__main__':
# dataroot = '/data2/3rscan_raw'
dataroot = '/data2/bim'
output_folder = os.path.join(dataroot,'viz')
scans_folder = os.path.join(dataroot,'scans')
if '3rscan' in dataroot:
rgb_folder = 'sequence' # color
rgb_posfix = '.color.jpg' #'.jpg'
scans_folder = dataroot
elif 'sgslam' in dataroot:
rgb_folder = 'rgb'
rgb_posfix = '.png'
elif 'bim' in dataroot:
rgb_folder = 'color'
rgb_posfix = '.jpg'
scans_folder = os.path.join(dataroot, 'test')
else:
rgb_folder = 'color'
rgb_folder = '.jpg'
frame_gap = 10
sample_frame_number = 100000
split_file = 'test.txt'
visualize_mask = False
### original prediction are based on rotated rgb ###
prediction_folder = 'prediction_no_augment'
rotated = False
tmp_scan_list = [
'280d8ebb-6cc6-2788-9153-98959a2da801',
'4731976c-f9f7-2a1a-95cc-31c4d1751d0b',
'1d2f850c-d757-207c-8fba-60b90a7d4691',
'ea318260-0a4c-2749-9389-4c16c782c4b1',
'10b17957-3938-2467-88a5-9e9254930dad',
]
scans = read_scan_list(os.path.join(dataroot, 'splits', split_file))
pred_folders = [os.path.join(scans_folder, scan, prediction_folder) for scan in scans]
# pred_folders = glob.glob(os.path.join(scans_folder, '*', prediction_folder))
print('find {} pred folders'.format(len(pred_folders)))
for pred_folder in pred_folders:
scene = pred_folder.split('/')[-2]
# scene_viz_folder = os.path.join(output_folder, scene)
scene_viz_folder = os.path.join(scans_folder, scene, 'pred_viz')
if os.path.exists(scene_viz_folder)==False:
os.makedirs(scene_viz_folder)
# if os.path.exists(scene_viz_folder): continue
if 'lg' in scene: continue
print('--------- processing {}----------'.format(scene))
pred_files = glob.glob(os.path.join(pred_folder, '*.json'))
if len(pred_files)<=sample_frame_number:
sample_pred_files = pred_files
else:
sample_pred_files = np.random.choice(pred_files, sample_frame_number, replace=False)
count = 0
for frame_pred in sorted(sample_pred_files):
frame_name = frame_pred.split('/')[-1][:12]
frame_id = int(frame_name[6:])
if count % 50==0:
print('{} / {}'.format(frame_name, len(sample_pred_files)))
if True:
# print(frame_name)
if os.path.exists(scene_viz_folder) == False:
os.makedirs(scene_viz_folder)
color_file = os.path.join(scans_folder, scene, rgb_folder, frame_name+rgb_posfix)
rgb = cv2.imread(color_file)
rgb = cv2.cvtColor(rgb, cv2.COLOR_BGR2RGB)
if rotated:
rgb = cv2.rotate(rgb, cv2.ROTATE_90_CLOCKWISE)
boxes, semantics, tags = load_pred(frame_pred)
for box, semantic_label_dict in zip(boxes, semantics):
label = list(semantic_label_dict.keys())[0]
score = semantic_label_dict[label]
label_score_str = '{}({:.2f})'.format(label, score)
rgb = draw_object_detection(rgb, box.numpy().astype(np.int32).reshape(-1),
label_score_str,
color=(0,255,0), thickness=2)
# print(label, score)
cv2.imwrite(os.path.join(scene_viz_folder, frame_name+'.jpg'), rgb)
count +=1
continue
if len(boxes)<1:continue
plt.figure(figsize=(10, 10))
plt.imshow(rgb)
for box, token_pair in zip(boxes, semantics):
box_vec = np.array(box).reshape(-1) # (4,)
if np.sum(box_vec)>1e-3:
show_box_tokens(box_vec, plt.gca(), token_pair)
if visualize_mask:
mask = cv2.imread(os.path.join(pred_folder, frame_name+'_mask.png'), cv2.IMREAD_UNCHANGED)
for i in range(1, mask.max()+1):
mask_i = (mask==i).astype(np.uint8)
show_mask(mask_i, plt.gca(), random_color=True)
plt.title(tags)
plt.savefig(os.path.join(scene_viz_folder, frame_name+'.jpg'))
count +=1
break
# break
print('{} saved {} frames'.format(scene, count))
# break