-
Notifications
You must be signed in to change notification settings - Fork 11
/
Copy pathdetect_rec_img.py
184 lines (173 loc) · 9.92 KB
/
detect_rec_img.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
import cv2
import numpy as np
import argparse
from itertools import product as product
from math import ceil
from PIL import Image, ImageDraw, ImageFont
CHARS = ['京', '沪', '津', '渝', '冀', '晋', '蒙', '辽', '吉', '黑',
'苏', '浙', '皖', '闽', '赣', '鲁', '豫', '鄂', '湘', '粤',
'桂', '琼', '川', '贵', '云', '藏', '陕', '甘', '青', '宁',
'新',
'0', '1', '2', '3', '4', '5', '6', '7', '8', '9',
'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'J', 'K',
'L', 'M', 'N', 'P', 'Q', 'R', 'S', 'T', 'U', 'V',
'W', 'X', 'Y', 'Z', 'I', 'O', '-'
]
def puttext_chinese(img, text, point, color):
pilimg = Image.fromarray(img) ###[:,:,::-1] BGRtoRGB
draw = ImageDraw.Draw(pilimg) # 图片上打印汉字
fontsize = int(min(img.shape[:2]) * 0.04)
font = ImageFont.truetype("simhei.ttf", fontsize, encoding="utf-8")
draw.text(point, text, color, font=font)
img = np.asarray(pilimg) ###[:,:,::-1] BGRtoRGB
return img
class detect_plate_recognition:
def __init__(self, confidence_threshold=0.9, top_k=5000, nms_threshold=0.4, keep_top_k=750, vis_thres=0.6):
self.model = cv2.dnn.readNet('mnet_plate.onnx')
self.im_height = 640
self.im_width = 640
self.scale = np.array([[self.im_width, self.im_height]], dtype=np.float32)
self.confidence_threshold = confidence_threshold
self.top_k = top_k
self.nms_threshold = nms_threshold
self.keep_top_k = keep_top_k
self.vis_thres = vis_thres
self.min_sizes = [[24, 48], [96, 192], [384, 768]]
self.steps = [8, 16, 32]
self.variance = [0.1, 0.2]
self.clip = False
self.prior_data = self.generate_priors() ####PriorBox生成的一堆anchor在强项推理过程中始终是常数是不变量，因此只需要在构造函数里定义一次即可
self.points_ref = np.float32([[0, 0], [94, 0], [0, 24], [94, 24]])
self.LPR = lprnet()
def generate_priors(self):
feature_maps = [[ceil(self.im_height / step), ceil(self.im_width / step)] for step in self.steps]
anchors = []
for k, f in enumerate(feature_maps):
min_sizes = self.min_sizes[k]
for i, j in product(range(f[0]), range(f[1])):
for min_size in min_sizes:
s_kx = min_size / self.im_width
s_ky = min_size / self.im_height
dense_cx = [x * self.steps[k] / self.im_width for x in [j + 0.5]]
dense_cy = [y * self.steps[k] / self.im_height for y in [i + 0.5]]
for cy, cx in product(dense_cy, dense_cx):
anchors += [cx, cy, s_kx, s_ky]
# back to torch land
output = np.asarray(anchors).reshape(-1, 4)
if self.clip:
output = np.clip(output, 0, 1)
return output
def decode(self, loc, priors, variances):
boxes = np.concatenate((priors[:, :2] + loc[:, :2] * variances[0] * priors[:, 2:], priors[:, 2:] * np.exp(loc[:, 2:] * variances[1])), axis=1)
boxes[:, :2] -= boxes[:, 2:] / 2
# boxes[:, 2:] += boxes[:, :2]
return boxes ### xmin,ymin, width, height
def decode_landm(self, pre, priors, variances):
landms = np.concatenate((priors[:, :2] + pre[:, :2] * variances[0] * priors[:, 2:],
priors[:, :2] + pre[:, 2:4] * variances[0] * priors[:, 2:],
priors[:, :2] + pre[:, 4:6] * variances[0] * priors[:, 2:],
priors[:, :2] + pre[:, 6:8] * variances[0] * priors[:, 2:],
), axis=1)
return landms
def resize_image(self, srcimg):
top, left, newh, neww = 0, 0, self.im_height, self.im_width
if srcimg.shape[0] != srcimg.shape[1]:
hw_scale = srcimg.shape[0] / srcimg.shape[1]
if hw_scale > 1:
newh, neww = self.im_height, int(self.im_width / hw_scale)
img = cv2.resize(srcimg, (neww, newh), interpolation=cv2.INTER_AREA)
left = int((self.im_width - neww) * 0.5)
img = cv2.copyMakeBorder(img, 0, 0, left, self.im_width - neww - left, cv2.BORDER_CONSTANT, value=0) # add border
else:
newh, neww = int(self.im_height * hw_scale), self.im_width
img = cv2.resize(srcimg, (neww, newh), interpolation=cv2.INTER_AREA)
top = int((self.im_height - newh) * 0.5)
img = cv2.copyMakeBorder(img, top, self.im_height - newh - top, 0, 0, cv2.BORDER_CONSTANT, value=0)
else:
img = cv2.resize(srcimg, (self.im_width, self.im_height), interpolation=cv2.INTER_AREA)
return img, newh, neww, top, left
def detect_rec(self, srcimg):
img, newh, neww, top, left = self.resize_image(srcimg)
blob = cv2.dnn.blobFromImage(img, mean=(104, 117, 123))
self.model.setInput(blob)
loc, conf, landms = self.model.forward(['loc', 'conf', 'landms'])
boxes = self.decode(loc, self.prior_data, self.variance)
boxes = boxes * np.tile(self.scale, (1, 2)) ####广播法则
scores = conf[:, 1]
landms = self.decode_landm(landms, self.prior_data, self.variance)
landms = landms * np.tile(self.scale, (1, 4)) ####广播法则
# ignore low scores
inds = np.where(scores > self.confidence_threshold)[0]
boxes = boxes[inds]
scores = scores[inds]
landms = landms[inds]
indices = cv2.dnn.NMSBoxes(boxes.tolist(), scores.tolist(), self.confidence_threshold, self.nms_threshold, top_k=self.keep_top_k)
boxes -= np.array([[left,top,0,0]]) ###还原到原图上，合理使用广播法则
landms -= np.tile(np.array([[left,top]]), (1, 4)) ####4个关键点坐标是x1,y1,x2,y2,x3,y3,x4,y4排列
srcim_scale = np.array([[srcimg.shape[1]/neww, srcimg.shape[0]/newh]], dtype=np.float32)
boxes = boxes * np.tile(srcim_scale, (1, 2)) ###还原到原图上，合理使用广播法则
landms = landms * np.tile(srcim_scale, (1, 4)) ####4个关键点坐标是x1,y1,x2,y2,x3,y3,x4,y4排列
# boxes = boxes.astype(np.int32)
# landms = landms.astype(np.int32)
for i in indices:
idx = i[0]
if scores[idx]<self.vis_thres:
continue
xmin, ymin, width, height = boxes[idx, :]
new_x1, new_y1 = landms[idx, 4] - xmin, landms[idx, 5]- ymin
new_x2, new_y2 = landms[idx, 6] - xmin, landms[idx, 7] - ymin
new_x3, new_y3 = landms[idx, 2] - xmin, landms[idx, 3] - ymin
new_x4, new_y4 = landms[idx, 0] - xmin, landms[idx, 1] - ymin
# 定义对应的点
points = np.float32([[new_x1, new_y1], [new_x2, new_y2], [new_x3, new_y3], [new_x4, new_y4]])
M = cv2.getPerspectiveTransform(points, self.points_ref)
img_box = srcimg[int(ymin):int(ymin + height), int(xmin):int(xmin + width), :]
processed = cv2.warpPerspective(img_box, M, (94, 24))
result = self.LPR.rec(processed)
# cv2.imshow('plate', processed)
cv2.rectangle(srcimg, (int(xmin), int(ymin)), (int(xmin + width), int(ymin + height)), (0, 0, 255), thickness=1)
# cv2.putText(srcimg, str(round(scores[idx], 3)), (int(xmin), int(ymin) - 5), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), thickness=2)
for j in range(4):
cv2.circle(srcimg, (int(landms[idx, 2*j]), int(landms[idx, 2*j+1])), 10, (255, 0, 0), thickness=-1)
srcimg = puttext_chinese(srcimg, result, (int(xmin), int(ymin) - 30), (0, 255, 0))
return srcimg
class lprnet:
def __init__(self):
self.img_size = (94, 24) ###width, height
self.model = cv2.dnn.readNet('Final_LPRNet_model.onnx')
def rec(self, img):
blob = cv2.dnn.blobFromImage(img, scalefactor=1 / 128, size=self.img_size, mean=127.5)
self.model.setInput(blob)
preb = self.model.forward()
# preb_label = []
# for j in range(preb.shape[1]):
# preb_label.append(np.argmax(preb[:, j], axis=0))
preb_label = np.argmax(preb, axis=0)
no_repeat_blank_label = []
pre_c = preb_label[0]
if pre_c != len(CHARS) - 1:
no_repeat_blank_label.append(pre_c)
for c in preb_label: # dropout repeate label and blank label
if (pre_c == c) or (c == len(CHARS) - 1):
if c == len(CHARS) - 1:
pre_c = c
continue
no_repeat_blank_label.append(c)
pre_c = c
return ''.join(list(map(lambda x: CHARS[x], no_repeat_blank_label)))
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='RetinaPL')
parser.add_argument('--imgpath', default='/home/wangbo/Desktop/data/yolo/license-plate-detect-recoginition/License-Plate-Detector-master/imgs/3.jpg', type=str, help='show detection results')
parser.add_argument('--confidence_threshold', default=0.02, type=float, help='confidence_threshold')
parser.add_argument('--top_k', default=1000, type=int, help='top_k')
parser.add_argument('--nms_threshold', default=0.4, type=float, help='nms_threshold')
parser.add_argument('--keep_top_k', default=500, type=int, help='keep_top_k')
parser.add_argument('--vis_thres', default=0.6, type=float, help='visualization_threshold')
args = parser.parse_args()
net = detect_plate_recognition(confidence_threshold=args.confidence_threshold, top_k=args.top_k, nms_threshold=args.nms_threshold, keep_top_k=args.keep_top_k, vis_thres=args.vis_thres)
srcimg = cv2.imread(args.imgpath)
srcimg = net.detect_rec(srcimg)
cv2.namedWindow('image', cv2.WINDOW_NORMAL)
cv2.imshow('image', srcimg)
cv2.waitKey(0)
cv2.destroyAllWindows()