Skip to content

affective_computing / mobile_face_recognize

Go to a project
Toggle navigation
Toggle navigation pinning
d77bff18 by jiangwenqiang
Options

remove q

1 parent f6b88746
Inline Side-by-side
Showing 1 changed file with 0 additions and 220 deletions
q deleted 100644 → 0
1 import os
2 import numpy as np
3 import MNN
4 import cv2
5 import logging
6 from retinaface import PriorBox
7
8
9 def py_cpu_nms(dets, thresh):
10 """Pure Python NMS baseline."""
11 x1 = dets[:, 0]
12 y1 = dets[:, 1]
13 x2 = dets[:, 2]
14 y2 = dets[:, 3]
15 scores = dets[:, 4]
16
17 areas = (x2 - x1 + 1) * (y2 - y1 + 1)
18 order = scores.argsort()[::-1]
19
20 keep = []
21 while order.size > 0:
22 i = order[0]
23 keep.append(i)
24 xx1 = np.maximum(x1[i], x1[order[1:]])
25 yy1 = np.maximum(y1[i], y1[order[1:]])
26 xx2 = np.minimum(x2[i], x2[order[1:]])
27 yy2 = np.minimum(y2[i], y2[order[1:]])
28
29 w = np.maximum(0.0, xx2 - xx1 + 1)
30 h = np.maximum(0.0, yy2 - yy1 + 1)
31 inter = w * h
32 ovr = inter / (areas[i] + areas[order[1:]] - inter)
33
34 inds = np.where(ovr <= thresh)[0]
35 order = order[inds + 1]
36
37 return keep
38
39
40 def decode_landm(pre, priors, variances):
41
42 landms = np.concatenate((priors[:, :2] + pre[:, :2] * variances[0] * priors[:, 2:],
43 priors[:, :2] + pre[:, 2:4] * variances[0] * priors[:, 2:],
44 priors[:, :2] + pre[:, 4:6] * variances[0] * priors[:, 2:],
45 priors[:, :2] + pre[:, 6:8] * variances[0] * priors[:, 2:],
46 priors[:, :2] + pre[:, 8:10] * variances[0] * priors[:, 2:]), 1)
47 return landms
48
49
50 def decode(loc, priors, variances):
51
52 boxes = np.concatenate((
53 priors[:, :2] + loc[:, :2] * variances[0] * priors[:, 2:],
54 priors[:, 2:] * np.exp(loc[:, 2:] * variances[1])), 1)
55 boxes[:, :2] -= boxes[:, 2:] / 2
56 boxes[:, 2:] += boxes[:, :2]
57 return boxes
58
59
60 class Face_Detector(object):
61 def __init__(self, model_path):
62 logging.info('******** Start Init Face Detector ********')
63 self.det_interpreter = MNN.Interpreter(model_path)
64 self.det_session = self.det_interpreter.createSession()
65 self.det_input_tensor = self.det_interpreter.getSessionInput(self.det_session)
66 logging.info('******** Success Init Face Detector ********')
67
68 def detect(self, frame, thr):
69 logging.info('******** Start Face Detect ********')
70 input_size = 320
71 img = cv2.resize(frame, (input_size, input_size))
72
73 img = np.float32(img)
74 im_height, im_width, _ = img.shape
75 scale = np.array([img.shape[1], img.shape[0], img.shape[1], img.shape[0]])
76 scale1 = np.array([img.shape[1], img.shape[0], img.shape[1], img.shape[0],
77 img.shape[1], img.shape[0], img.shape[1], img.shape[0],
78 img.shape[1], img.shape[0]])
79
80 w_r = input_size/frame.shape[1]
81 h_r = input_size/frame.shape[0]
82
83 confidence_threshold = 0.02
84 vis_threshold = 0.5
85 nms_threshold = 0.4
86 keep_top_k = 100
87 variance = [0.1, 0.2]
88 img -= (104, 117, 123)
89 img = img.transpose(2, 0, 1)
90 img = np.expand_dims(img, axis=0)
91
92 input_tensor = MNN.Tensor((1, 3, input_size, input_size), MNN.Halide_Type_Float, img, MNN.Tensor_DimensionType_Caffe)
93 self.det_input_tensor.copyFrom(input_tensor)
94 self.det_interpreter.runSession(self.det_session)
95
96 bbox_output_tensor = self.det_interpreter.getSessionOutput(self.det_session, 'output0')
97 conf_output_tensor = self.det_interpreter.getSessionOutput(self.det_session, 'output1')
98 landmark_output_tensor = self.det_interpreter.getSessionOutput(self.det_session, 'output2')
99
100 bbox_output = bbox_output_tensor.getData()
101 conf_output = conf_output_tensor.getData()
102 landmark_output = landmark_output_tensor.getData()
103
104 norm_confs = list()
105 for i in range(int(len(conf_output)/2)):
106 norm_confs.append([conf_output[i * 2 + 0], conf_output[i * 2 + 1]])
107
108 norm_bboxes = list()
109 for i in range(int(len(conf_output)/2)):
110 norm_bboxes.append([bbox_output[i * 4 + 0], bbox_output[i * 4 + 1], bbox_output[i * 4 + 2], bbox_output[i * 4 + 3]])
111
112 norm_landmarks = list()
113 for i in range(int(len(conf_output)/2)):
114 norm_landmarks.append([landmark_output[i * 10 + 0], landmark_output[i * 10 + 1],
115 landmark_output[i * 10 + 2], landmark_output[i * 10 + 3],
116 landmark_output[i * 10 + 4], landmark_output[i * 10 + 5],
117 landmark_output[i * 10 + 6], landmark_output[i * 10 + 7],
118 landmark_output[i * 10 + 8], landmark_output[i * 10 + 9]])
119
120 norm_confs = np.array(norm_confs)
121 norm_bboxes = np.array(norm_bboxes)
122 norm_landmarks = np.array(norm_landmarks)
123
124 priorbox = PriorBox(image_size=(im_height, im_width))
125 priors = priorbox.forward()
126
127 scores = norm_confs[:, 1]
128
129 boxes = decode(norm_bboxes, priors, variance)
130 boxes = boxes * scale
131
132 landms = decode_landm(norm_landmarks, priors, variance)
133 landms = landms * scale1
134
135 # ignore low scores
136 inds = np.where(scores > confidence_threshold)[0]
137 boxes = boxes[inds]
138 landms = landms[inds]
139 scores = scores[inds]
140
141 # keep top-K before NMS
142 order = scores.argsort()[::-1][:keep_top_k]
143 boxes = boxes[order]
144 landms = landms[order]
145 scores = scores[order]
146
147 # do NMS
148 dets = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False)
149 keep = py_cpu_nms(dets, nms_threshold)
150
151 dets = dets[keep, :]
152 landms = landms[keep]
153
154 # keep top-K faster NMS
155 dets = dets[:keep_top_k, :]
156 landms = landms[:keep_top_k, :]
157
158 dets = np.concatenate((dets, landms), axis=1)
159 face_bboxes = []
160 face_landmarks = []
161 max_area = float('-inf')
162 max_index = 0
163
164 i = 0
165 for b in dets:
166 if b[4] < vis_threshold:
167 continue
168 resize_b = []
169 x1 = int(b[0] / w_r)
170 y1 = int(b[1] / h_r)
171 x2 = int(b[2] / w_r)
172 y2 = int(b[3] / h_r)
173 x3 = int(b[5] / w_r)
174 y3 = int(b[6] / h_r)
175 x4 = int(b[7] / w_r)
176 y4 = int(b[8] / h_r)
177 x5 = int(b[9] / w_r)
178 y5 = int(b[10] / h_r)
179 x6 = int(b[11] / w_r)
180 y6 = int(b[12] / h_r)
181 x7 = int(b[13] / w_r)
182 y7 = int(b[14] / h_r)
183 resize_b = [x1, y1, x2, y2, 0, x3, y3, x4, y4, x5, y5, x6, y6, x7, y7]
184
185 # cv2.rectangle(frame, (resize_b[0], resize_b[1]), (resize_b[2], resize_b[3]), (0, 0, 255), 2)
186 # cv2.circle(frame, (resize_b[5], resize_b[6]), 1, (0, 0, 255), 4)
187 # cv2.circle(frame, (resize_b[7], resize_b[8]), 1, (0, 255, 255), 4)
188 # cv2.circle(frame, (resize_b[9], resize_b[10]), 1, (255, 0, 255), 4)
189 # cv2.circle(frame, (resize_b[11], resize_b[12]), 1, (0, 255, 0), 4)
190 # cv2.circle(frame, (resize_b[13], resize_b[14]), 1, (255, 0, 0), 4)
191
192 area = (resize_b[2] - resize_b[0]) * (resize_b[3] - resize_b[1])
193 if area > max_area:
194 max_area = area
195 max_index = i
196 i += 1
197
198 face_bboxes.append([resize_b[0], resize_b[1], resize_b[2], resize_b[3]])
199 face_landmarks.append([(resize_b[5], resize_b[6]),
200 (resize_b[7], resize_b[8]),
201 (resize_b[9], resize_b[10]),
202 (resize_b[11], resize_b[12]),
203 (resize_b[13], resize_b[14])])
204
205 # import time
206 # cv2.imwrite('results/0.jpg', frame)
207 return face_bboxes, face_landmarks, max_index
208
209
210 if __name__ == '__main__':
211 det_face_model_path = r'/home/jwq/PycharmProjects/situ/src/face_det/Pytorch_Retinaface/weights/mobilenet_0.25.mnn'
212 image_path = r'input/3.jpg'
213 image_save_path = r'results/3.jpg'
214 thr = 0.5
215
216 face_detector = Face_Detector(det_face_model_path)
217 image = cv2.imread(image_path)
218 face_detector.detect(image, thr)
219 # image_ploted = face_detector.plot(image, face_bboxes)
220 # cv2.imwrite(image_save_path, image_ploted)
Styling with Markdown is supported
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!