-
Notifications
You must be signed in to change notification settings - Fork 32
/
Copy pathmain.py
189 lines (156 loc) · 7.19 KB
/
main.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
import numpy as np
import cv2
from faster_rcnn_wrapper import FasterRCNNSlim
from _tf_compat_import import compat_tensorflow as tf
import argparse
import os
import json
import time
from nms_wrapper import NMSType, NMSWrapper
def detect(sess, rcnn_cls, image):
# pre-processing image for Faster-RCNN
img_origin = image.astype(np.float32, copy=True)
img_origin -= np.array([[[102.9801, 115.9465, 112.7717]]])
img_shape = img_origin.shape
img_size_min = np.min(img_shape[:2])
img_size_max = np.max(img_shape[:2])
img_scale = 600 / img_size_min
if np.round(img_scale * img_size_max) > 1000:
img_scale = 1000 / img_size_max
img = cv2.resize(img_origin, None, None, img_scale, img_scale, cv2.INTER_LINEAR)
img_info = np.array([img.shape[0], img.shape[1], img_scale], dtype=np.float32)
img = np.expand_dims(img, 0)
# test image
_, scores, bbox_pred, rois = rcnn_cls.test_image(sess, img, img_info)
# bbox transform
boxes = rois[:, 1:] / img_scale
boxes = boxes.astype(bbox_pred.dtype, copy=False)
widths = boxes[:, 2] - boxes[:, 0] + 1
heights = boxes[:, 3] - boxes[:, 1] + 1
ctr_x = boxes[:, 0] + 0.5 * widths
ctr_y = boxes[:, 1] + 0.5 * heights
dx = bbox_pred[:, 0::4]
dy = bbox_pred[:, 1::4]
dw = bbox_pred[:, 2::4]
dh = bbox_pred[:, 3::4]
pred_ctr_x = dx * widths[:, np.newaxis] + ctr_x[:, np.newaxis]
pred_ctr_y = dy * heights[:, np.newaxis] + ctr_y[:, np.newaxis]
pred_w = np.exp(dw) * widths[:, np.newaxis]
pred_h = np.exp(dh) * heights[:, np.newaxis]
pred_boxes = np.zeros_like(bbox_pred, dtype=bbox_pred.dtype)
pred_boxes[:, 0::4] = pred_ctr_x - 0.5 * pred_w
pred_boxes[:, 1::4] = pred_ctr_y - 0.5 * pred_h
pred_boxes[:, 2::4] = pred_ctr_x + 0.5 * pred_w
pred_boxes[:, 3::4] = pred_ctr_y + 0.5 * pred_h
# clipping edge
pred_boxes[:, 0::4] = np.maximum(pred_boxes[:, 0::4], 0)
pred_boxes[:, 1::4] = np.maximum(pred_boxes[:, 1::4], 0)
pred_boxes[:, 2::4] = np.minimum(pred_boxes[:, 2::4], img_shape[1] - 1)
pred_boxes[:, 3::4] = np.minimum(pred_boxes[:, 3::4], img_shape[0] - 1)
return scores, pred_boxes
def load_file_from_dir(dir_path):
ret = []
for file in os.listdir(dir_path):
path_comb = os.path.join(dir_path, file)
if os.path.isdir(path_comb):
ret += load_file_from_dir(path_comb)
else:
ret.append(path_comb)
return ret
def fmt_time(dtime):
if dtime <= 0:
return '0:00.000'
elif dtime < 60:
return '0:%02d.%03d' % (int(dtime), int(dtime * 1000) % 1000)
elif dtime < 3600:
return '%d:%02d.%03d' % (int(dtime / 60), int(dtime) % 60, int(dtime * 1000) % 1000)
else:
return '%d:%02d:%02d.%03d' % (int(dtime / 3600), int((dtime % 3600) / 60), int(dtime) % 60,
int(dtime * 1000) % 1000)
def main():
parser = argparse.ArgumentParser(description='Anime face detector demo')
parser.add_argument('-i', help='The input path of an image or directory', required=True, dest='input', type=str)
parser.add_argument('-o', help='The output json path of the detection result', dest='output')
parser.add_argument('-nms', help='Change the threshold for non maximum suppression',
dest='nms_thresh', default=0.3, type=float)
parser.add_argument('-conf', help='Change the threshold for class regression', dest='conf_thresh',
default=0.8, type=float)
parser.add_argument('-model', help='Specify a new path for model', dest='model', type=str,
default='model/res101_faster_rcnn_iter_60000.ckpt')
parser.add_argument('-nms-type', help='Type of nms', choices=['PY_NMS', 'CPU_NMS', 'GPU_NMS'], dest='nms_type',
default='CPU_NMS')
parser.add_argument('-crop-location', help='The output folder to place the cropped images', dest='crop_output_image_location')
parser.add_argument('-start-output', help='Start the numbering of the cropped images filename', dest='start_output_number',
default=0, type=int)
parser.add_argument('-crop-width', help='The width of images to crop', dest='crop_width', type=int)
parser.add_argument('-crop-height', help='The height of images to crop', dest='crop_height', type=int)
args = parser.parse_args()
assert os.path.exists(args.input), 'The input path does not exists'
if os.path.isdir(args.input):
files = load_file_from_dir(args.input)
else:
files = [args.input]
file_len = len(files)
if args.nms_type == 'PY_NMS':
nms_type = NMSType.PY_NMS
elif args.nms_type == 'CPU_NMS':
nms_type = NMSType.CPU_NMS
elif args.nms_type == 'GPU_NMS':
nms_type = NMSType.GPU_NMS
else:
raise ValueError('Incorrect NMS Type, not supported yet')
nms = NMSWrapper(nms_type)
cfg = tf.ConfigProto()
cfg.gpu_options.allow_growth = True
sess = tf.Session(config=cfg)
net = FasterRCNNSlim()
saver = tf.train.Saver()
saver.restore(sess, args.model)
result = {}
time_start = time.time()
for idx, file in enumerate(files):
elapsed = time.time() - time_start
eta = (file_len - idx) * elapsed / idx if idx > 0 else 0
print('[%d/%d] Elapsed: %s, ETA: %s >> %s' % (idx+1, file_len, fmt_time(elapsed), fmt_time(eta), file))
img = cv2.imread(file)
if img is None:
continue
scores, boxes = detect(sess, net, img)
boxes = boxes[:, 4:8]
scores = scores[:, 1]
keep = nms(np.hstack([boxes, scores[:, np.newaxis]]).astype(np.float32), args.nms_thresh)
boxes = boxes[keep, :]
scores = scores[keep]
inds = np.where(scores >= args.conf_thresh)[0]
scores = scores[inds]
boxes = boxes[inds, :]
result[file] = []
for i in range(scores.shape[0]):
x1, y1, x2, y2 = boxes[i, :].tolist()
new_result = {'score': float(scores[i]),
'bbox': [x1, y1, x2, y2]}
result[file].append(new_result)
if args.output is None and args.crop_output_image_location is None:
cv2.rectangle(img, (int(x1), int(y1)), (int(x2), int(y2)), (0, 0, 255), 2)
if args.crop_output_image_location:
cropped_image = img[int(y1):int(y2), int(x1):int(x2)]
if args.crop_width and args.crop_height:
cropped_image = cv2.resize(cropped_image,
(args.crop_width, args.crop_height),
interpolation = cv2.INTER_AREA)
cv2.imwrite(args.crop_output_image_location + str(args.start_output_number) + ".jpg", cropped_image)
args.start_output_number += 1
if args.output:
if ((idx+1) % 1000) == 0:
# saving the temporary result
with open(args.output, 'w') as f:
json.dump(result, f)
elif args.crop_output_image_location is None:
cv2.imshow(file, img)
if args.output:
with open(args.output, 'w') as f:
json.dump(result, f)
else:
cv2.waitKey()
if __name__ == '__main__':
main()