人脸检测业务理论及实现

 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

<annotaion>
	<floder>widerface</floder>
	<filename>0--Parade_0_Parade_marchingband_1_5.jpg</filename> #指向了JPEGImage下面的一个图片
	<source> #图片来源
		<databse>wider face Database</databse>
		<annotation>PASCAL VOC2007</annotaion>
		<image>flickr</image>
		<flickrid>1</flickerid>
	<source>
	<owner>
		<flickerid>ieblyang</flickerid>
		<name>ieblyang</name>
	</owner>
	<size> #当前图片的尺寸，通道数
		<width>1024</width>
		<height>683</height>
		<depth>3</depth>
	</size>
	<segmented>0</segmented>
	<object> #一个object对应一个人脸
		<name>face</name>
		<pose>Unspecified</poss>
		<truncated>1</truncated>
		<diffcult>0</diffcult>
		<bndbox>#记录了矩形框的坐标值
			<xmin>495</xmin>
			<ymin>347</ymin>
			<xmax>532</xmax>
			<ymax>398</ymax>
		</bndbox>
	</object>
...
</annotaion>		   

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

# wider_face_train_bbx_gt.txt
0--Parade/0_Parade_marchingband_1_849.jpg   # 图片路径
1                                           # 针对当前图片中的人脸个数
449 330 122 149 0 0 0 0 0 0                 # 当前人脸的标准信息
↑   ↑
前两个值人脸框左上角坐标
        ↑   ↑
        第三四个值表示人脸框的大小
                ↑ ↑ ↑ ↑ ↑ ↑
                后六个值分别表示Scale、Pose、Occlusion、Expression、Makeup、Illumation
                利用这六个值可以对人脸数据进行清洗

1
2
3
4
5
6
7
8

size = doc.createElement('size')		# 获取图片的基本信息
annotation.appendChild(size)
width = doc.createElement('width')	   # 宽度
width.appendChild(doc.createTextNode(str(saveimg.shape[1])))
height = doc.createElement('height')	# 高度
height.appendChild(doc.createTextNode(str(saveimg.shape[0])))
depth = doc.createElement('depth')		# 通道数量
depth.appendChild(doc.createTextNode(str(saveimg.shape[2])))

 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

# bboxes 当前图片中人脸框的坐标
for i in range(len(bboxes)):		
bbox = bboxes[i]
objects = doc.createElement('object')
annotation.appendChild(objects)
object_name = doc.createElement('name')
object_name.appendChild(doc.createTextNode('face'))
objects.appendChild(object_name)
pose = doc.createElement('pose')
pose.appendChild(doc.createTextNode('Unspecified'))
objects.appendChild(pose)
truncated = doc.createElement('truncated')
truncated.appendChild(doc.createTextNode('0'))
objects.appendChild(truncated)
difficult = doc.createElement('difficult')
difficult.appendChild(doc.createTextNode('0'))
objects.appendChild(difficult)
bndbox = doc.createElement('bndbox')
objects.appendChild(bndbox)
xmin = doc.createElement('xmin')
xmin.appendChild(doc.createTextNode(str(bbox[0])))
bndbox.appendChild(xmin)
ymin = doc.createElement('ymin')
ymin.appendChild(doc.createTextNode(str(bbox[1])))
bndbox.appendChild(ymin)
xmax = doc.createElement('xmax')
xmax.appendChild(doc.createTextNode(str(bbox[0] + bbox[2])))
bndbox.appendChild(xmax)
ymax = doc.createElement('ymax')
ymax.appendChild(doc.createTextNode(str(bbox[1] + bbox[3])))
bndbox.appendChild(ymax)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19

sc = max(im_data.shape)
im_data_tmp = numpy.zeros([sc, sc, 3],  dtype = numpy.uint8)
off_w = (sc - im_data.shape[1])  //  2
off_h = (sc - im_data.shape[0])  //  2
# 对图片进行周围填充，填充为正方形
im_data_tmp[off_h:im_data.shape[0] + off_h, off_w:im_data.shape[1] + off_w, ...] = im_data
im_data = im_data_tmp
numbox = int(gt.readline())
bboxes = []
for i in range(numbox):
line = gt.readline()
infos = line.split(" ")
for j in range(infos.__len__() - 1):
infos[j] = int(infos[j])
# 数据清洗：保留resize到640×640 尺寸在8×8以上的人脸
if infos[2] * 80 < im_data.shape[1] or infos[3] * 80 < im_data.shape[0]:
continue
bbox = (infos[0] + off_w, infos[1] + off_h, infos[2], infos[3])
bboxes.append(bbox)

1
2
3
4
5

python3 object_detection/dataset_tools/create_face_tf_record.py \
--data_dir=/mnt/dataset \
--year=widerface \
--output_path=/mnt/dataset/widerface/TF-data/train.record \
--set=train

1
2
3
4
5
6

python3 object_detection/model_main.py \ 
--pipeline_config_path= \
/mnt/FaceAI/models/research/object_detection/samples/configs/ssd_resnet50_v1_fpn_shared_box_predictor_640x640_coco14_sync_face.config 
--model_dir=/mnt/dataset/widerface/resnet50v1-fpn \
--num_train_steps=100000 \
--alsologtostder \

1
2
3
4
5
6
7
8

python3 object_detection/export_inference_graph.py \ 
--input_type =image_tensor \
--popeline_config_path= \
/mnt/FaceAI/models/research/object_detection/samples/configs/ssd_resnet50_v1_fpn_shared_box_predictor_640x640_coco14_sync_face.config \
--trained_checkpoint_prefix= \
/mnt/FaceAI/dataset/widerface/model/resnet50v1-fpn/model.ckpt-75065
--output_directory= \
/mnt/FaceAI/dataset/widerface/model/resnet50v1-fpn/pb

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15

for image_path in im_path_list:
imdata = cv2.imread(image_path)
sp = imdata.shape
imdata = cv2.resize(imdata, IMAGE_SIZE)
output_dict  =  run_inference_for_single_image( imdata,  detection_graph )
for i in range(len(output_dict['detection_scores'])):
if output_dict['detection_scores'][i] > 0.6:
    bbox = output_dict['detection_scores'][i]
    y1 = int(IMAGE_SIZE[0] * bbox[0])
    x1 = int(IMAGE_SIZE[1] * bbox[1])
    y2 = int(IMAGE_SIZE[0] * bbox[2])
    x2 = int(IMAGE_SIZE[1] * bbox[3])
    cv2.rectangle(imdata, (x1,y1), (x2,y2), (0,255,0), 2)
cv2.imshow("im", imdata)
cv2.waitKey(0)