forked from yjh0410/YOLO-Nano
-
Notifications
You must be signed in to change notification settings - Fork 0
/
test.py
233 lines (200 loc) · 7.44 KB
/
test.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
import argparse
import cv2
import os
import time
import numpy as np
import torch
import torch.backends.cudnn as cudnn
from data.voc import VOC_CLASSES, VOCDetection
from data.coco import coco_class_index, coco_class_labels, COCODataset
from data import config
from data.transforms import ValTransforms
from utils.misc import TestTimeAugmentation
parser = argparse.ArgumentParser(description='YOLO-Nano Detection')
# basic
parser.add_argument('-size', '--img_size', default=416, type=int,
help='img_size')
parser.add_argument('--show', action='store_true', default=False,
help='show the visulization results.')
parser.add_argument('-vs', '--visual_threshold', default=0.3, type=float,
help='Final confidence threshold')
parser.add_argument('--cuda', action='store_true', default=False,
help='use cuda.')
parser.add_argument('--save_folder', default='det_results/', type=str,
help='Dir to save results')
# model
parser.add_argument('-v', '--version', default='yolo_nano',
help='yolo_nano')
parser.add_argument('--trained_model', default='weight/',
type=str, help='Trained state_dict file path to open')
parser.add_argument('--conf_thresh', default=0.1, type=float,
help='NMS threshold')
parser.add_argument('--nms_thresh', default=0.45, type=float,
help='NMS threshold')
# dataset
parser.add_argument('--root', default='/mnt/share/ssd2/dataset',
help='data root')
parser.add_argument('-d', '--dataset', default='coco',
help='coco.')
# TTA
parser.add_argument('-tta', '--test_aug', action='store_true', default=False,
help='use test augmentation.')
args = parser.parse_args()
def plot_bbox_labels(img, bbox, label=None, cls_color=None, text_scale=0.4):
x1, y1, x2, y2 = bbox
x1, y1, x2, y2 = int(x1), int(y1), int(x2), int(y2)
t_size = cv2.getTextSize(label, 0, fontScale=1, thickness=2)[0]
# plot bbox
cv2.rectangle(img, (x1, y1), (x2, y2), cls_color, 2)
if label is not None:
# plot title bbox
cv2.rectangle(img, (x1, y1-t_size[1]), (int(x1 + t_size[0] * text_scale), y1), cls_color, -1)
# put the test on the title bbox
cv2.putText(img, label, (int(x1), int(y1 - 5)), 0, text_scale, (0, 0, 0), 1, lineType=cv2.LINE_AA)
return img
def visualize(img,
bboxes,
scores,
cls_inds,
vis_thresh,
class_colors,
class_names,
class_indexs=None,
dataset_name='voc'):
ts = 0.4
for i, bbox in enumerate(bboxes):
if scores[i] > vis_thresh:
cls_id = int(cls_inds[i])
if dataset_name == 'coco':
cls_color = class_colors[cls_id]
cls_id = class_indexs[cls_id]
else:
cls_color = class_colors[cls_id]
if len(class_names) > 1:
mess = '%s: %.2f' % (class_names[cls_id], scores[i])
else:
cls_color = [255, 0, 0]
mess = None
img = plot_bbox_labels(img, bbox, mess, cls_color, text_scale=ts)
return img
def test(args,
net,
device,
dataset,
transforms=None,
vis_thresh=0.4,
class_colors=None,
class_names=None,
class_indexs=None,
show=False,
test_aug=None,
dataset_name='coco'):
num_images = len(dataset)
save_path = os.path.join('det_results/', args.dataset, args.version)
os.makedirs(save_path, exist_ok=True)
for index in range(num_images):
print('Testing image {:d}/{:d}....'.format(index+1, num_images))
image, _ = dataset.pull_image(index)
h, w, _ = image.shape
size = np.array([[w, h, w, h]])
# prepare
x, _, _, scale, offset = transforms(image)
x = x.unsqueeze(0).to(device)
t0 = time.time()
# forward
# test augmentation:
if test_aug is not None:
bboxes, scores, cls_inds = test_aug(x, net)
else:
# inference
bboxes, scores, cls_inds = net(x)
print("detection time used ", time.time() - t0, "s")
# rescale
bboxes -= offset
bboxes /= scale
bboxes *= size
# vis detection
img_processed = visualize(
img=image,
bboxes=bboxes,
scores=scores,
cls_inds=cls_inds,
vis_thresh=vis_thresh,
class_colors=class_colors,
class_names=class_names,
class_indexs=class_indexs,
dataset_name=dataset_name
)
if show:
cv2.imshow('detection', img_processed)
cv2.waitKey(0)
# save result
cv2.imwrite(os.path.join(save_path, str(index).zfill(6) +'.jpg'), img_processed)
if __name__ == '__main__':
args = parser.parse_args()
# cuda
if args.cuda:
print('use cuda')
cudnn.benchmark = True
device = torch.device("cuda")
else:
device = torch.device("cpu")
# dataset and evaluator
if args.dataset == 'voc':
data_dir = os.path.join(args.root, 'VOCdevkit')
class_names = VOC_CLASSES
class_indexs = None
num_classes = 20
anchor_size = config.MULTI_ANCHOR_SIZE
dataset = VOCDetection(
data_dir=data_dir,
img_size=args.img_size)
elif args.dataset == 'coco':
data_dir = os.path.join(args.root, 'COCO')
class_names = coco_class_labels
class_indexs = coco_class_index
num_classes = 80
anchor_size = config.MULTI_ANCHOR_SIZE_COCO
dataset = COCODataset(
data_dir=data_dir,
img_size=args.img_size,
image_set='val2017')
else:
print('unknow dataset !! Only support voc and coco !!')
exit(0)
np.random.seed(0)
class_colors = [(np.random.randint(255),
np.random.randint(255),
np.random.randint(255)) for _ in range(num_classes)]
# build model
if args.version == 'yolo_nano':
from models.yolo_nano import YOLONano
backbone = '1.0x'
net = YOLONano(device=device,
input_size=args.img_size,
num_classes=num_classes,
anchor_size=anchor_size,
backbone=backbone)
print('Let us train yolo_nano ......')
else:
print('Unknown version !!!')
exit()
# load weight
net.load_state_dict(torch.load(args.trained_model, map_location=device))
net.to(device).eval()
print('Finished loading model!')
# TTA
test_aug = TestTimeAugmentation(num_classes=num_classes) if args.test_aug else None
# run
test(args=args,
net=net,
device=device,
dataset=dataset,
transforms=ValTransforms(args.img_size),
vis_thresh=args.visual_threshold,
class_colors=class_colors,
class_names=class_names,
class_indexs=class_indexs,
show=args.show,
test_aug=test_aug,
dataset_name=args.dataset)