def test(stage, testFolder):
print("Start testing in %s"%(testFolder))
detectors = [None, None, None]
if stage in ['pnet', 'rnet', 'onet']:
modelPath = os.path.join(rootPath, 'tmp/model/pnet/')
a = [b[5:-6] for b in os.listdir(modelPath) if b.startswith('pnet-') and b.endswith('.index')]
maxEpoch = max(map(int, a)) # auto match a max epoch model
modelPath = os.path.join(modelPath, "pnet-%d"%(maxEpoch))
print("Use PNet model: %s"%(modelPath))
detectors[0] = FcnDetector(P_Net,modelPath)
if stage in ['rnet', 'onet']:
modelPath = os.path.join(rootPath, 'tmp/model/rnet/')
a = [b[5:-6] for b in os.listdir(modelPath) if b.startswith('rnet-') and b.endswith('.index')]
maxEpoch = max(map(int, a))
modelPath = os.path.join(modelPath, "rnet-%d"%(maxEpoch))
print("Use RNet model: %s"%(modelPath))
detectors[1] = Detector(R_Net, 24, 1, modelPath)
if stage in ['onet']:
modelPath = os.path.join(rootPath, 'tmp/model/onet/')
a = [b[5:-6] for b in os.listdir(modelPath) if b.startswith('onet-') and b.endswith('.index')]
maxEpoch = max(map(int, a))
modelPath = os.path.join(modelPath, "onet-%d"%(maxEpoch))
print("Use ONet model: %s"%(modelPath))
detectors[2] = Detector(O_Net, 48, 1, modelPath)
mtcnnDetector = MtcnnDetector(detectors=detectors, min_face_size = 24, threshold=[0.9, 0.6, 0.7])
testImages = []
for name in os.listdir(testFolder):
testImages.append(os.path.join(testFolder, name))
testDatas = TestLoader(testImages)
# Now to detect
allBoxes, allLandmarks = mtcnnDetector.detect_face(testDatas)
print("\n")
# Save it
for idx, imagePath in enumerate(testImages):
savePath = os.path.join(rootPath, 'testing', 'results_%s'%(stage))
if not os.path.isdir(savePath):
os.makedirs(savePath)
image = cv2.imread(imagePath)
save_bboxes(savePath,idx,image,allBoxes[idx])
for bbox in allBoxes[idx]:
cv2.putText(image,str(np.round(bbox[4],2)),(int(bbox[0]),int(bbox[1])),cv2.FONT_HERSHEY_TRIPLEX,1,color=(255,0,255))
cv2.rectangle(image, (int(bbox[0]),int(bbox[1])),(int(bbox[2]),int(bbox[3])),(0,0,255))
allLandmark = allLandmarks[idx]
if allLandmark is not None: # pnet and rnet will be ignore landmark
for landmark in allLandmark:
for i in range(len(landmark)/2):
cv2.circle(image, (int(landmark[2*i]),int(int(landmark[2*i+1]))), 3, (0,0,255))
cv2.imwrite(os.path.join(savePath, "result_%d.jpg" %(idx)), image)
print("Save image to %s"%(savePath))
def main(args):
'''通过PNet或RNet生成下一个网络的输入'''
size = args.input_size
batch_size = config.batches
min_face_size = config.min_face
stride = config.stride
thresh = config.thresh
#模型地址
model_path = ['../model/PNet/', '../model/RNet/', '../model/ONet']
if size == 12:
net = 'PNet'
save_size = 24
elif size == 24:
net = 'RNet'
save_size = 48
# 图片数据地址
base_dir = 'g:/mtcnn-dataset/data/WIDER_train/'
# 处理后的图片存放地址
data_dir = 'g:/mtcnn-dataset/data/%d' % (save_size)
neg_dir = os.path.join(data_dir, 'negative')
pos_dir = os.path.join(data_dir, 'positive')
part_dir = os.path.join(data_dir, 'part')
for dir_path in [neg_dir, pos_dir, part_dir]:
if not os.path.exists(dir_path):
os.makedirs(dir_path)
detectors = [None, None, None]
PNet = FcnDetector(P_Net, model_path[0])
detectors[0] = PNet
if net == 'RNet':
RNet = Detector(R_Net, 24, batch_size[1], model_path[1])
detectors[1] = RNet
filename = '../data/wider_face_train_bbx_gt.txt'
#读取文件的image和box对应函数在utils中
data = read_annotation(base_dir, filename)
mtcnn_detector = MtcnnDetector(detectors,
min_face_size=min_face_size,
stride=stride,
threshold=thresh)
save_path = data_dir
save_file = os.path.join(save_path, 'detections.pkl')
if not os.path.exists(save_file):
#将data制作成迭代器
print('载入数据')
test_data = TestLoader(data['images'])
detectors, _ = mtcnn_detector.detect_face(test_data)
print('完成识别')
with open(save_file, 'wb') as f:
pickle.dump(detectors, f, 1)
print('开始生成图像')
save_hard_example(save_size, data, neg_dir, pos_dir, part_dir, save_path)
def test(stage, testFolder):
print("Start testing in %s"%(testFolder))
detectors = [None, None, None]
if stage in ['pnet', 'rnet', 'onet']:
modelPath = os.path.join(rootPath, 'tmp/model/pnet/')
a = [b[5:-6] for b in os.listdir(modelPath) if b.startswith('pnet-') and b.endswith('.index')]
maxEpoch = max(map(int, a)) # auto match a max epoch model
modelPath = os.path.join(modelPath, "pnet-%d"%(maxEpoch))
print("Use PNet model: %s"%(modelPath))
detectors[0] = FcnDetector(P_Net,modelPath)
if stage in ['rnet', 'onet']:
modelPath = os.path.join(rootPath, 'tmp/rnet/model/middle/')
a = [b[5:-6] for b in os.listdir(modelPath) if b.startswith('rnet-') and b.endswith('.index')]
maxEpoch = max(map(int, a))
modelPath = os.path.join(modelPath, "rnet-%d"%(maxEpoch))
print("Use RNet model: %s"%(modelPath))
detectors[1] = Detector(R_Net, 24, 1, modelPath)
if stage in ['onet']:
modelPath = os.path.join(rootPath, 'tmp/onet/model/small/')
a = [b[5:-6] for b in os.listdir(modelPath) if b.startswith('onet-') and b.endswith('.index')]
maxEpoch = max(map(int, a))
modelPath = os.path.join(modelPath, "onet-%d"%(maxEpoch))
print("Use ONet model: %s"%(modelPath))
detectors[2] = Detector(O_Net, 48, 1, modelPath)
mtcnnDetector = MtcnnDetector(detectors=detectors, min_face_size =12, threshold=[0.6, 0.6, 0.7],scale_factor=0.7)
testImages = []
for name in os.listdir(testFolder):
testImages.append(os.path.join(testFolder, name))
print("\n")
right_num=0
miss_num=0
FN=0
# Save it
for idx, imagePath in enumerate(testImages):
if(idx<=6000):
image = cv2.imread(imagePath)
image = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
print(right_num,FN,miss_num)
try:
allBoxes, allLandmarks = mtcnnDetector.detect_face([image])
if(allBoxes.__len__()==1):
right_num+=1
else:
FN+=(allBoxes.__len__()-1)
except:
miss_num+=1
pass
else:
break
def test(stage):
detectors = [None, None, None]
if stage in ['pnet', 'rnet', 'onet']:
modelPath = os.path.join(rootPath, 'tmp/model/pnet/')
a = [b[5:-6] for b in os.listdir(modelPath) if b.startswith('pnet-') and b.endswith('.index')]
maxEpoch = max(map(int, a)) # auto match a max epoch model
modelPath = os.path.join(modelPath, "pnet-%d"%(maxEpoch))
print("Use PNet model: %s"%(modelPath))
detectors[0] = FcnDetector(P_Net,modelPath)
if stage in ['rnet', 'onet']:
modelPath = os.path.join(rootPath, 'tmp/model/rnet/')
a = [b[5:-6] for b in os.listdir(modelPath) if b.startswith('rnet-') and b.endswith('.index')]
maxEpoch = max(map(int, a))
modelPath = os.path.join(modelPath, "rnet-%d"%(maxEpoch))
print("Use RNet model: %s"%(modelPath))
detectors[1] = Detector(R_Net, 24, 1, modelPath)
if stage in ['onet']:
modelPath = os.path.join(rootPath, 'tmp/model/onet/')
a = [b[5:-6] for b in os.listdir(modelPath) if b.startswith('onet-') and b.endswith('.index')]
maxEpoch = max(map(int, a))
modelPath = os.path.join(modelPath, "onet-%d"%(maxEpoch))
print("Use ONet model: %s"%(modelPath))
detectors[2] = Detector(O_Net, 48, 1, modelPath)
mtcnnDetector = MtcnnDetector(detectors=detectors, min_face_size =50, threshold=[0.8, 0.8, 0.9],scale_factor=0.4)
cap = cv2.VideoCapture(0)
while(True):
testImages = []
ret, image = cap.read()
image=cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
testImages.append(image)
# Now to detect
starttime=time.time()
allBoxes, allLandmarks = mtcnnDetector.detect_face(testImages)
# print("\n")
# Save it
# print(time.time()-starttime)
for bbox in allBoxes[0]:
cv2.putText(image,str(np.round(bbox[4],2)),(int(bbox[0]),int(bbox[1])),cv2.FONT_HERSHEY_TRIPLEX,1,color=(255,0,255))
cv2.rectangle(image, (int(bbox[0]),int(bbox[1])),(int(bbox[2]),int(bbox[3])),(0,0,255))
allLandmark = allLandmarks[0]
if allLandmark is not None: # pnet and rnet will be ignore landmark
for landmark in allLandmark:
for i in range(int(len(landmark)/2)):
cv2.circle(image, (int(landmark[2*i]),int(int(landmark[2*i+1]))), 3, (0,0,255))
cv2.imshow("test", image)
c = cv2.waitKey(1) & 0xFF
if c == 27 or c == ord('q'):
break
def test_net(batch_size, stage, thresh, min_face_size, stride):
print(">>>>>> Detect bbox for %s..." % (stage))
detectors = [None, None, None]
if stage in ["rnet", "onet"]:
modelPath = os.path.join(rootPath, 'tmp/model/pnet/')
# 第几个checkpoint
a = [
b[5:-6] for b in os.listdir(modelPath)
if b.startswith('pnet-') and b.endswith('.index')
]
maxEpoch = max(map(int, a))
modelPath = os.path.join(modelPath, "pnet-%d" % (maxEpoch))
print("Use PNet model: %s" % (modelPath))
PNet = FcnDetector(P_Net, modelPath)
detectors[0] = PNet
if stage in ["onet"]:
modelPath = os.path.join(rootPath, 'tmp/model/rnet/')
a = [
b[5:-6] for b in os.listdir(modelPath)
if b.startswith('rnet-') and b.endswith('.index')
]
maxEpoch = max(map(int, a))
modelPath = os.path.join(modelPath, "rnet-%d" % (maxEpoch))
print("Use RNet model: %s" % (modelPath))
RNet = Detector(R_Net, 24, batch_size, modelPath)
detectors[1] = RNet
# read annatation(type:dict)
widerImagesPath = os.path.join(rootPath, "dataset", "WIDER_train",
"images")
annoTxtPath = os.path.join(rootPath, "dataset",
"wider_face_train_bbx_gt.txt")
# data['images'], data['bboxes']
data = read_wider_annotation(widerImagesPath, annoTxtPath)
mtcnn_detector = MtcnnDetector(detectors=detectors,
min_face_size=min_face_size,
stride=stride,
threshold=thresh)
test_data = TestLoader(data['images'])
# do detect
detections, _ = mtcnn_detector.detect_face(test_data)
# save detect result
save_path = os.path.join(rootPath, "tmp/data", stage)
if not os.path.exists(save_path):
os.makedirs(save_path)
save_file = os.path.join(save_path, "detections.pkl")
with open(save_file, 'wb') as f:
pickle.dump(detections, f, 1)
print("\nDone! Start to do OHEM...")
__save_data(stage, data, save_path)
class face_detect(object):
def __init__(self):
self.load_model()
def load_model(self):
thresh = [0.6, 0.7, 0.7]
min_face_size = 20
stride = 2
slide_window = False
detectors = [None, None, None]
prefix = [
'./weight/PNet_landmark/PNet', './weight/RNet_landmark/RNet',
'./weight/ONet_landmark/ONet'
]
epoch = [18, 14, 16]
model_path = ['%s-%s' % (x, y) for x, y in zip(prefix, epoch)]
PNet, RNet, ONet = FcnDetector(P_Net, model_path[0]), Detector(R_Net, 24, 1, model_path[1]), \
Detector(O_Net, 48, 1, model_path[2])
detectors[0], detectors[1], detectors[2] = PNet, RNet, ONet
self.mtcnn_detector = MtcnnDetector(detectors=detectors,
min_face_size=min_face_size,
stride=stride,
threshold=thresh,
slide_window=slide_window)
def detect(self, img, show_pic=False):
"""
:param img: BGR格式的图片
:param show_pic: 是否展示图片
:return:boxes_c type is ndarray, shape is (M, 5)
"""
boxes_c, landmarks = self.mtcnn_detector.detect(img)
if show_pic:
for i in range(boxes_c.shape[0]):
bbox = boxes_c[i, :4]
score = boxes_c[i, 4]
corpbbox = [
int(bbox[0]),
int(bbox[1]),
int(bbox[2]),
int(bbox[3])
]
cv2.rectangle(frame, (corpbbox[0], corpbbox[1]),
(corpbbox[2], corpbbox[3]), (255, 0, 0), 2)
cv2.putText(frame, '{:.3f}'.format(score),
(corpbbox[0], corpbbox[1] - 2),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 3)
for i in range(landmarks.shape[0]):
for j in range(int(len(landmarks[i]) / 2)):
cv2.circle(frame, (int(landmarks[i][2 * j]),
int(int(landmarks[i][2 * j + 1]))), 2,
(0, 0, 255))
cv2.putText(frame, 'handsome men', (30, 30),
cv2.FONT_HERSHEY_SIMPLEX, 1, (127, 255, 212), 2)
cv2.imshow("", frame)
if cv2.waitKey(0) & 0xFF == ord('q'):
cv2.destroyAllWindows()
return boxes_c
def load_model(self):
thresh = [0.6, 0.7, 0.7]
min_face_size = 20
stride = 2
slide_window = False
detectors = [None, None, None]
prefix = [
'./weight/PNet_landmark/PNet', './weight/RNet_landmark/RNet',
'./weight/ONet_landmark/ONet'
]
epoch = [18, 14, 16]
model_path = ['%s-%s' % (x, y) for x, y in zip(prefix, epoch)]
PNet, RNet, ONet = FcnDetector(P_Net, model_path[0]), Detector(R_Net, 24, 1, model_path[1]), \
Detector(O_Net, 48, 1, model_path[2])
detectors[0], detectors[1], detectors[2] = PNet, RNet, ONet
self.mtcnn_detector = MtcnnDetector(detectors=detectors,
min_face_size=min_face_size,
stride=stride,
threshold=thresh,
slide_window=slide_window)
def realse(stack):
print('Begin to get frame......')
os.environ["CUDA_VISIBLE_DEVICES"] = '1'
detectors = net('onet')
mtcnnDetector = MtcnnDetector(detectors=detectors,
min_face_size=24,
threshold=[0.9, 0.6, 0.7])
while True:
if len(stack) > 30:
image = stack.pop()
image = cv2.resize(image,
(int(image.shape[1]), int(image.shape[0])))
image = np.array(image)
boxes_c, _ = mtcnnDetector.detect_video(image)
for bbox in boxes_c:
x1 = int(bbox[0])
y1 = int(bbox[1])
x2 = int(bbox[2])
y2 = int(bbox[3])
cv2.putText(image,
str(np.round(bbox[4], 2)),
(int(bbox[0]), int(bbox[1])),
cv2.FONT_HERSHEY_TRIPLEX,
0.3,
color=(0, 255, 0))
cv2.rectangle(image, (x1, y1), (x2, y2), (0, 0, 255))
# cut = image[y1:y2, x1:x2]
# for i in range(len(boxes_c)):
# cv2.imwrite(str(i) + '.jpg', cut)
print('deteced face: ({},{}), ({},{})'.format(x1, y1, x2, y2))
cv2.imshow("Detected", image)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cv2.destroyAllWindows()
def load_mtcnn():
MODEL_PATH = config.MTCNN_MODEL_PATH
MIN_FACE_SIZE = int(config.MIN_FACE_SIZE)
STEPS_THRESHOLD = [float(i) for i in config.STEPS_THRESHOLD.split(",")]
detectors = [None, None, None]
prefix = [MODEL_PATH + "/PNet_landmark/PNet",
MODEL_PATH + "/RNet_landmark/RNet",
MODEL_PATH + "/ONet_landmark/ONet"]
epoch = [18, 14, 16]
model_path = ['%s-%s' % (x, y) for x, y in zip(prefix, epoch)]
PNet = FcnDetector(P_Net, model_path[0])
detectors[0] = PNet
RNet = Detector(R_Net, 24, 1, model_path[1])
detectors[1] = RNet
ONet = Detector(O_Net, 48, 1, model_path[2])
detectors[2] = ONet
mtcnn_detector = MtcnnDetector(detectors=detectors, min_face_size=MIN_FACE_SIZE, threshold=STEPS_THRESHOLD)
return mtcnn_detector
model_path = ['model/PNet/', 'model/RNet/', 'model/ONet']
batch_size = config.batches
PNet = PDetector(P_Net, model_path[0])
detectors[0] = PNet
#判断最后测试选择的网络
if test_mode in ["RNet", "ONet"]:
RNet = Detector(R_Net, 24, batch_size[1], model_path[1])
detectors[1] = RNet
if test_mode == "ONet":
ONet = Detector(O_Net, 48, batch_size[2], model_path[2])
detectors[2] = ONet
mtcnn_detector = MtcnnDetector(detectors=detectors,
min_face_size=min_face_size,
stride=stride,
threshold=thresh)
#设定输出路径
out_path = config.out_path
#两种输入方式，1为图片，2为摄像头
if config.input_mode == '1':
#设定测试图片的路径
path = config.test_dir
#print(path)
for item in os.listdir(path):
img_path = os.path.join(path, item)
img = cv2.imread(img_path)
#获知图片中的人脸框和关键点
boxes_c, landmarks = mtcnn_detector.detect(img)
for i in range(boxes_c.shape[0]):
bbox = boxes_c[i, :4]
def test(stage, testFolder):
print("Start testing in %s" % (testFolder))
detectors = [None, None, None]
if stage in ['pnet', 'rnet', 'onet']:
modelPath = os.path.join(rootPath, 'tmp/origin/model/pnet')
a = [
b[5:-6] for b in os.listdir(modelPath)
if b.startswith('pnet-') and b.endswith('.index')
]
maxEpoch = max(map(int, a)) # auto match a max epoch model
modelPath = os.path.join(modelPath, "pnet-%d" % (maxEpoch))
print("Use PNet model: %s" % (modelPath))
detectors[0] = FcnDetector(P_Net, modelPath)
if stage in ['rnet', 'onet']:
modelPath = os.path.join(rootPath, 'tmp/origin/model/rnet')
a = [
b[5:-6] for b in os.listdir(modelPath)
if b.startswith('rnet-') and b.endswith('.index')
]
maxEpoch = max(map(int, a))
modelPath = os.path.join(modelPath, "rnet-%d" % (maxEpoch))
print("Use RNet model: %s" % (modelPath))
detectors[1] = Detector(R_Net, 24, 1, modelPath)
if stage in ['onet']:
modelPath = os.path.join(rootPath, 'tmp/origin/model/onet')
a = [
b[5:-6] for b in os.listdir(modelPath)
if b.startswith('onet-') and b.endswith('.index')
]
maxEpoch = max(map(int, a))
modelPath = os.path.join(modelPath, "onet-%d" % (maxEpoch))
print("Use ONet model: %s" % (modelPath))
detectors[2] = Detector(O_Net, 48, 1, modelPath)
mtcnnDetector = MtcnnDetector(detectors=detectors,
min_face_size=50,
threshold=[0.8, 0.9, 0.9])
fileFoldName = "faceListInt.txt"
outFilename = 'F:/software/yansan/MTCNN-on-FDDB-Dataset-master/FDDB-folds/' + 'predict.txt' # fileOutName
foldFilename = 'F:/software/yansan/MTCNN-on-FDDB-Dataset-master/FDDB-folds/' + fileFoldName
prefixFilename = 'E:/database/FDDB_Face Detection Data Set and Benchmark/'
fout = open(outFilename, 'a+')
f = open(foldFilename, 'r') # FDDB-fold-00.txt, read
for imgpath in tqdm(f.readlines()):
testImages = []
imgpath = imgpath.split('\n')[0]
# foutOnce.write(imgpath+'\n')
# foutFold.write(imgpath+'\r')
img = cv2.imread(prefixFilename + imgpath + '.jpg')
img = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
if img is None:
continue
testImages.append(img)
boundingboxes, points = mtcnnDetector.detect_face(testImages)
# boundingboxes, points = demo.detect_face(img_matlab, minsize, PNet, RNet, ONet, threshold, False, factor)
text1 = str(imgpath) + '\n' + str(len(boundingboxes[0])) + '\n'
fout.write(text1) # FDDB-fold-%02d-out.txt or predict.txt
for bbox in boundingboxes[0]:
# print(bbox,"???")
text2 = str(int(bbox[0])) + ' ' + str(int(bbox[1])) + ' ' \
+ str(abs(int(bbox[2] - bbox[0]))) + ' ' \
+ str(abs(int(bbox[3] - bbox[1]))) + ' ' \
+ str(bbox[4]) + '\n'
fout.write(text2) # FDDB-fold-%02d-out.txt or predict.txt
# text2 = str(int(boundingboxes[coordinate][0][0]))
# fout.write(text2) # FDDB-fold-%02d-out.txt or predict.txt
# print error
f.close() # input the fold list, FDDB-fold-00.txt
fout.close() # output the result, predict.txt
PNet = PDetector(P_Net, 12, batch_size[0], model_path[0])
else:
PNet = PDetector(P_Net, model_path[0])
detectors[0] = PNet
# load rnet model
RNet = RODetector(R_Net, 24, batch_size[1], model_path[1])
detectors[1] = RNet
# load onet model
ONet = Detector(O_Net, 48, batch_size[2], model_path[2])
detectors[2] = ONet
mtcnn_detector = MtcnnDetector(detectors=detectors,
min_face_size=min_face_size,
stride=stride,
threshold=thresh,
slide_window=slide_window)
gt_imdb = []
path = "xixi"
for item in os.listdir(path):
gt_imdb.append(os.path.join(path, item))
test_data = TestLoader(gt_imdb)
all_boxes, landmarks = mtcnn_detector.detect_face(test_data)
count = 0
for imagepath in gt_imdb:
image = cv2.imread(imagepath)
model_path = ['model/PNet/', 'model/RNet/', 'model/ONet']
batch_size = [2048, 256, 32]
PNet = FcnDetector(P_Net, model_path[0]) # detecotors for PNet
detectors[0] = PNet
# in and output path
path = 'picture'
out_path = 'output'
detectors[1] = Detector(R_Net, 24, batch_size[1], model_path[1])
detectors[2] = Detector(O_Net, 48, batch_size[2], model_path[2])
# Use the three detectors to construct a
mtcnn_detector = MtcnnDetector(detectors=detectors,
min_face_size=min_face_size,
stride=stride,
threshold=thresh,
scale_factor=0.909)
#%%
#%%
def box_of_landmarks(bboxes, landmark):
"""
pick the box contains all the landmarks in bboxes
:bboxes: all bboxes, must be shape of (_, 5), sorted descending by probility
:landmarks:
"""
assert landmark.shape == (10, ), 'error landmark shape'
assert (bboxes.shape[0] > 0) & (bboxes.shape[1] == 5), 'error bboxes shape'
# 模型放置位置
model_path = ['model/PNet/', 'model/RNet/', 'model/ONet']
batch_size = config.batches
detectors[0] = FcnDetector(P_Net, model_path[0]) # detecotors for PNet
if test_mode in ['RNet', 'ONet']:
detectors[1] = Detector(R_Net, 24, batch_size[1], model_path[1])
if test_mode == 'ONet':
detectors[2] = Detector(O_Net, 48, batch_size[2], model_path[2])
# Use the three detectors to construct a
mtcnn_detector = MtcnnDetector(detectors=detectors,
min_face_size=min_face_size,
stride=stride,
threshold=thresh,
scale_factor=scale_factor)
start_path = '/home/sherk/Workspace/mtcnn-pytorch/'
filenames = os.listdir(start_path + 'img')
missing_detection = 0
false_detection = 0
all_detection = 0
all_labels = 0
for filename in tqdm(filenames):
iou_threshold = 0.4
image = start_path + 'img/{}'.format(filename)
# 模型放置位置
model_path = ['model/PNet/', 'model/RNet/', 'model/ONet']
batch_size = config.batches
PNet = FcnDetector(P_Net, model_path[0])
detectors[0] = PNet
if test_mode in ["RNet", "ONet"]:
RNet = Detector(R_Net, 24, batch_size[1], model_path[1])
detectors[1] = RNet
if test_mode == "ONet":
ONet = Detector(O_Net, 48, batch_size[2], model_path[2])
detectors[2] = ONet
mtcnn_detector = MtcnnDetector(detectors=detectors,
min_face_size=min_face_size,
stride=stride,
threshold=thresh)
def oneImg(path):
img = cv2.imread(path)
boxes_c, landmarks = mtcnn_detector.detect(img)
for i in range(boxes_c.shape[0]):
bbox = boxes_c[i, :4]
score = boxes_c[i, 4]
corpbbox = [int(bbox[0]), int(bbox[1]), int(bbox[2]), int(bbox[3])]
# 画人脸框
cv2.rectangle(img, (corpbbox[0], corpbbox[1]),
(corpbbox[2], corpbbox[3]), (255, 0, 0), 1)
# 判别为人脸的置信度
cv2.putText(img, '{:.2f}'.format(score),
cv2.putText(image,
str(name), (left, top - 10),
cv2.FONT_HERSHEY_TRIPLEX,
1,
color=(0, 255, 0))
cv2.rectangle(image, (left, top), (right, bottom), (0, 0, 255), 2)
cv2.imwrite(str(img_path) + '.jpg', image)
cv2.imshow('Recognition_Results', image)
cv2.waitKey(500)
cv2.destroyAllWindows()
if __name__ == '__main__':
logset()
detectors = net('onet')
os.environ["CUDA_VISIBLE_DEVICES"] = '1'
mtcnnDetector = MtcnnDetector(detectors=detectors,
min_face_size=40,
threshold=[0.9, 0.6, 0.7])
for image_file in os.listdir("/home/lxz/project/faceid/datasets/test"):
full_file_path = os.path.join("/home/lxz/project/faceid/datasets/test",
image_file)
predictions = predict(
full_file_path,
model_path="/home/lxz/project/faceid/LFW_classifier_model.clf")
show_results(
os.path.join("/home/lxz/project/faceid/datasets/test", image_file),
predictions)
def recognize(stack):
logging.info("[INFO]:Starting video stream...")
# os.environ["CUDA_VISIBLE_DEVICES"] = '1'
# config = tf.ConfigProto()
# config.gpu_options.allow_growth = True
# session = tf.Session(config = config)
data = pickle.loads(
open('/home/lxz/project/faceid/alignment.pickle', "rb").read())
detectors = net('onet')
mtcnnDetector = MtcnnDetector(detectors=detectors,
min_face_size=60,
threshold=[0.9, 0.6, 0.7])
logging.info('MTCNN/KNN Model load sucessed !!!!')
while True:
if len(stack) > 20:
boxes = []
frame = stack.pop()
image = np.array(frame)
allBoxes, _ = mtcnnDetector.detect_video(image)
for box in allBoxes:
x_1 = int(box[0])
y_1 = int(box[1])
x_2 = int(box[2])
y_2 = int(box[3])
boxes.append((y_1 - 10, x_2 + 12, y_2 + 10, x_1 - 12))
logging.debug(boxes)
start = time.time()
# num_jitters（re-sample人脸的次数）参数的设定，数值越大精度相对会高，但是速度会慢；
encodings = face_recognition.face_encodings(frame,
boxes,
num_jitters=6)
end = time.time()
logging.info('[INFO]:Encoding face costed: {} s'.format(end -
start))
print('encode time is {}ms'.format((end - start) * 1000))
names = []
for encoding in encodings:
# distance between faces to consider it a match, optimize is 0.6
matches = face_recognition.compare_faces(data['encodings'],
encoding,
tolerance=0.35)
name = 'Stranger'
if True in matches:
matchesidx = [i for (i, b) in enumerate(matches) if b]
counts = {}
for i in matchesidx:
name = data['names'][i]
counts[name] = counts.get(name, 0) + 1
name = max(counts, key=counts.get)
logging.debug(name)
names.append(name)
# 绘制检测框 + 人脸识别结果
for ((top, right, bottom, left), name) in zip(boxes, names):
# print(name)
y1 = int(top)
x1 = int(right)
y2 = int(bottom)
x2 = int(left)
cv2.rectangle(frame, (x2, y1), (x1, y2), (0, 0, 255), 2)
if name == 'Stranger':
cv2.putText(frame, name, (x2, y1 - 5),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
else:
print(name)
cv2.putText(frame, name, (x2, y1 - 5),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 255, 0), 2)
cv2.imshow('Recognize-no-alignment', frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cv2.destroyAllWindows()
detectors = [None, None, None]
prefix = [
'./weight/PNet_landmark/PNet', './weight/RNet_landmark/RNet',
'./weight/ONet_landmark/ONet'
]
epoch = [18, 14, 16]
model_path = ['%s-%s' % (x, y) for x, y in zip(prefix, epoch)]
PNet = FcnDetector(P_Net, model_path[0])
detectors[0] = PNet
RNet = Detector(R_Net, 24, 1, model_path[1])
detectors[1] = RNet
ONet = Detector(O_Net, 48, 1, model_path[2])
detectors[2] = ONet
mtcnn_detector = MtcnnDetector(detectors=detectors,
min_face_size=min_face_size,
stride=stride,
threshold=thresh,
slide_window=slide_window)
video_capture = cv2.VideoCapture(0)
video_capture.set(3, 500)
video_capture.set(4, 600)
corpbbox = None
while True:
# fps = video_capture.get(cv2.CAP_PROP_FPS)
t1 = cv2.getTickCount()
ret, frame = video_capture.read()
if ret:
image = np.array(frame)
boxes_c, landmarks = mtcnn_detector.detect(image)
def test(stage, profiling):
print("Start Detecting")
detectors = [None, None, None]
if stage in ['pnet', 'rnet', 'onet']:
modelPath = os.path.join(rootPath, 'tmp/model/pnet/')
a = [
b[5:-6] for b in os.listdir(modelPath)
if b.startswith('pnet-') and b.endswith('.index')
]
maxEpoch = max(map(int, a)) # auto match a max epoch model
modelPath = os.path.join(modelPath, "pnet-%d" % (maxEpoch))
print("Use PNet model: %s" % (modelPath))
detectors[0] = FcnDetector(P_Net, modelPath, profiling)
if stage in ['rnet', 'onet']:
modelPath = os.path.join(rootPath, 'tmp/model/rnet/')
a = [
b[5:-6] for b in os.listdir(modelPath)
if b.startswith('rnet-') and b.endswith('.index')
]
maxEpoch = max(map(int, a))
modelPath = os.path.join(modelPath, "rnet-%d" % (maxEpoch))
print("Use RNet model: %s" % (modelPath))
detectors[1] = Detector(R_Net, 24, 1, modelPath, profiling)
if stage in ['onet']:
modelPath = os.path.join(rootPath, 'tmp/model/onet/')
a = [
b[5:-6] for b in os.listdir(modelPath)
if b.startswith('onet-') and b.endswith('.index')
]
maxEpoch = max(map(int, a))
modelPath = os.path.join(modelPath, "onet-%d" % (maxEpoch))
print("Use ONet model: %s" % (modelPath))
detectors[2] = Detector(O_Net, 48, 1, modelPath, profiling)
mtcnnDetector = MtcnnDetector(detectors=detectors,
min_face_size=24,
threshold=[0.9, 0.6, 0.7])
# Now to detect
camID = 0
cap = cv2.VideoCapture(camID)
while True:
ret, image = cap.read()
if ret == 0:
break
[h, w] = image.shape[:2]
print(h, w)
#image_data = cv2.flip(image, 1)
#image_data = cv2.flip(image, 1)
image_data = image
start_time = time.time()
testImages = []
testImages.append(image_data)
allBoxes, allLandmarks = mtcnnDetector.detect_face(testImages)
inf_time = time.time() - start_time
print("inference time(s): {}".format(inf_time))
del testImages[0]
#print("allBoxes: {}".format(allBoxes))
#print("allLandmarks: {}".format(allLandmarks))
#print("\n")
# Save it
if (len(allBoxes) >= 1):
for idx, bbox in enumerate(allBoxes):
cv2.putText(image_data,
str(np.round(bbox[idx][4], 2)),
(int(bbox[idx][0]), int(bbox[idx][1])),
cv2.FONT_HERSHEY_TRIPLEX,
1,
color=(255, 0, 255))
cv2.rectangle(image_data,
(int(bbox[idx][0]), int(bbox[idx][1])),
(int(bbox[idx][2]), int(bbox[idx][3])),
(0, 0, 255))
allLandmark = allLandmarks[idx][0].tolist()
total_landmark_pts = len(allLandmark)
if allLandmark is not None and len(
allLandmark
) == 10: # pnet and rnet will be ignore landmark
for index, landmark in enumerate(allLandmark):
for i in range(int(total_landmark_pts / 2)):
cv2.circle(image_data,
(int(allLandmark[2 * i]),
int(int(allLandmark[2 * i + 1]))), 3,
(255, 255, 255))
cv2.imshow('Face/Landmark Detection', image_data)
k = cv2.waitKey(1) & 0xff
if k == ord('q') or k == 27:
break
cap.release()
def realse(stack):
# 解决GPU内存占用问题
import tensorflow as tf
os.environ["CUDA_VISIBLE_DEVICES"] = '0'
config = tf.ConfigProto()
#config.gpu_options.per_process_gpu_memory_fraction = 0.3
config.gpu_options.allow_growth = True
session = tf.Session(config=config)
print('Begin to get frame......')
max_cosine_distance = 0.3
nn_budget = None
nms_max_overlap = 1.0
# deep_sort
model_filename = 'mars-small128.pb'
encoder = gdet.create_box_encoder(model_filename, batch_size=1)
metric = nn_matching.NearestNeighborDistanceMetric("cosine",
max_cosine_distance,
nn_budget)
tracker = Tracker(metric)
# mtcnn
detectors = mtcnn('onet')
mtcnnDetector = MtcnnDetector(detectors=detectors,
min_face_size=24,
threshold=[0.9, 0.6, 0.7])
while True:
if len(stack) >= 20:
frame = stack.pop()
#frame = cv2.resize(frame, (int(frame.shape[1]/3), int(frame.shape[0]/3)))
frame = np.array(frame)
# 输出tmp信息为[x,y,w,h]; x:检测框左上角点的x坐标；y:检测框左上角y坐标；w:框宽;h:框高；
# 原MTCNN输出的信息为检测框一对角点的坐标信息（左上角点、右下角点），以及检测为人脸的概率值[x1,y2,x2,y2,置信度]；
tmp, _ = mtcnnDetector.detect_follow(frame)
features = encoder(frame, tmp)
detections = [
Detection(bbox, 1.0, feature)
for bbox, feature in zip(tmp, features)
]
# Run non-maxima suppression.
boxes = np.array([d.tlwh for d in detections])
scores = np.array([d.confidence for d in detections])
indices = preprocessing.non_max_suppression(
boxes, nms_max_overlap, scores)
detections = [detections[i] for i in indices]
# Call the tracker
tracker.predict()
tracker.update(detections)
for track in tracker.tracks:
if not track.is_confirmed() or track.time_since_update > 1:
continue
bbox = track.to_tlbr()
cv2.rectangle(frame, (int(bbox[0]), int(bbox[1])),
(int(bbox[2]), int(bbox[3])), (0, 255, 0), 1)
cv2.putText(frame, str(track.track_id),
(int(bbox[0]), int(bbox[1])), 0, 0.5, (0, 255, 0),
2)
for det in detections:
bbox = det.to_tlbr()
cv2.rectangle(frame, (int(bbox[0]), int(bbox[1])),
(int(bbox[2]), int(bbox[3])), (0, 0, 255), 1)
cv2.imshow("Detected", frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cv2.destroyAllWindows()
