def detect(image_path, trackerid, ts, cameraId, face_filter):
result = m.detect(image_path)
#FIXME:
result = result.replace('[,', '[')
result = json.loads(result)
#print('detect result-----',result)
people_cnt = 0
cropped = []
detected = False
nrof_faces, img_data, imgs_style, blury_arr, face_width, face_height = load_align_image_v2(result, image_path, trackerid, ts, cameraId, face_filter)
if img_data is not None and len(img_data) > 0:
people_cnt = len(img_data)
detected = True
for align_image_path, prewhitened in img_data.items():
style=imgs_style[align_image_path]
blury=blury_arr[align_image_path]
width=face_width[align_image_path]
height=face_height[align_image_path]
cropped.append({"path": align_image_path, "style": style, "blury": blury, "ts": ts,
"trackerid": trackerid, "totalPeople": people_cnt, "cameraId": cameraId,
"width":width,"height":height})
return json.dumps({'detected': detected, "ts": ts, "totalPeople": people_cnt, "cropped": cropped, 'totalmtcnn': nrof_faces})
def onPhoto():
#test1 = imagen en formato base64 de alejandro
#test2 = imagen en formateo base54 de obama
img = b64ToImg(open("test_2", "rb").read())
id = detect(img)
if id is None:
print "Nada"
else:
print id
def demo(in_fn, out_fn):
print ">>> Loading image..."
img_color = cv2.imread(in_fn)
img_gray = cv2.cvtColor(img_color, cv.CV_RGB2GRAY)
img_gray = cv2.equalizeHist(img_gray)
print in_fn, img_gray.shape
print ">>> Detecting faces..."
start = time.time()
rects = detect(img_gray)
end = time.time()
print 'time:', end - start
img_out = img_color.copy()
draw_rects(img_out, rects, (0, 255, 0))
cv2.imwrite(out_fn, img_out)
def __init__(self,
img,
image_src='/tmp/temp.jpg',
segment_dump="/tmp/",
a_score=1,
prediction_stage=False,
min_saliency=0.02,
max_iter_slic=100):
self.timer = time.time()
self.image_src = image_src
self.a_score = a_score
self.image = img
# self.image = io.imread(image_src)
print "Image source : ", image_src
self.__set_timer("segmentation...")
segment_object = seg.SuperPixelSegmentation(self.image,
max_iter=max_iter_slic)
self.segment_map = segment_object.getSegmentationMap()
self.slic_map = segment_object.getSlicSegmentationMap()
self.__print_timer("segmentation")
self.__set_timer("saliency...")
saliency_object = saliency.Saliency(self.image, 3)
self.saliency_map = saliency_object.getSaliencyMap()
self.__print_timer("saliency")
# perform face detection
self.__set_timer("face detection...")
self.faces = face_detection.detect(np.array(self.image))
self.__print_timer("face detection")
self.__set_timer("saliency detection of objects...")
self.saliency_list, self.salient_objects, self.pixel_count, self.segment_map2 = cutils.detect_saliency_of_segments(
self.segment_map.astype(np.intp), self.saliency_map, min_saliency)
self.__print_timer("saliency detection of objects")
def clicked2(self):
self.label_3.setText("Status: Id/Roll No-" + self.lineEdit.text() +
", Name- " + self.lineEdit_2.text())
self.update()
face_detection.detect(self.lineEdit.text(), self.lineEdit_2.text())
'cpus': 2,
'result': 0
}, {
'res': 1080,
'file': './1_1920x1080.jpg',
'minsize': 200,
'cpus': 2,
'result': 0
}]
m.init('./model/')
print('warming up')
m.set_minsize(40)
m.set_num_threads(1)
m.set_threshold(0.6, 0.7, 0.8)
result = m.detect('./1_854x480.jpg')
print(result)
m.detect('./1_854x480.jpg')
m.detect('./1_854x480.jpg')
m.detect('./1_854x480.jpg')
m.detect('./1_854x480.jpg')
print('starting up')
rounds = 20
for item in benchmark:
print(item)
m.set_minsize(item['minsize'])
m.set_num_threads(item['cpus'])
start = time.time()
for i in range(rounds):
import face_detection
import face_recognition
import numpy as np
import pickle
import os
# image = 'images/thang.jpg'
# faces = face_detection.detect(image)
# face_recognition.learn(faces, 'thang')
folder = 'test'
for image in os.listdir(folder):
path = '/'.join([folder, image])
faces = face_detection.detect(path)
faces_with_name = face_recognition.recognition(faces, path)
import face_detection as m
import time
m.init('./models/ncnn/')
print('warming up')
m.set_minsize(40)
m.set_threshold(0.6, 0.7, 0.8)
m.set_num_threads(1)
m.detect('./images_480p/1_854x480.jpg')
m.detect('./images_480p/1_854x480.jpg')
m.detect('./images_480p/1_854x480.jpg')
m.detect('./images_480p/1_854x480.jpg')
m.detect('./images_480p/1_854x480.jpg')
start = time.time()
for i in range(100):
step_start = time.time()
result = m.detect('./images_480p/1_854x480.jpg')
step_end = time.time()
print('step {} duration is {}'.format(i, step_end - step_start))
end = time.time()
print(result)
print('average duration is {}'.format((end - start) / 100))
model_name = 'blind_with_regularization.model'
COM = 'COM9'
camera = 1
baudrate = 9600
width = 64
height = 64
prob = 0
label = ''
print("loading model .....")
model = load_model(model_name)
print("model loaded")
ard = Arduino(baudrate, COM) ##movleft(),movright()
vce = Voice() #left(),right()
st = Stop()
fac = detect()
current = datetime.datetime.now()
flag = None
cap = cv2.VideoCapture(camera)
ret = True
prev = None
while ret:
ret, frame = cap.read()
frame = cv2.resize(frame, (640, 480))
faces = fac.faceDetect(frame)
##stop on left
## you have a stop on '''
current = datetime.datetime.now()
import face_detection as m
import time
m.init('./model/')
m.set_minsize(100)
m.set_threshold(0.6, 0.7, 0.8)
m.set_num_threads(2)
print('warming up')
result = m.detect('./1_1920x1080.jpg')
print(result)
m.detect('./1_1920x1080.jpg')
m.detect('./1_1920x1080.jpg')
m.detect('./1_1920x1080.jpg')
m.detect('./1_1920x1080.jpg')
start = time.time()
for i in range(10):
step_start = time.time()
result = m.detect('./1_1920x1080.jpg')
step_end = time.time()
print('step {} duration is {}'.format(i, step_end - step_start))
end = time.time()
print(result)
print('1080p average duration is {}'.format((end - start) / 10))
