def generate(names, targets):
ret, frame = vid_source.read()
# tensor code
while ret:
img_bg = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
image = Image.fromarray(img_bg)
try:
bboxes, faces = face_recognize.align_multi(image)
except:
bboxes = []
faces = []
if len(bboxes) != 0:
bboxes = bboxes[:,:-1] #shape:[10,4],only keep 10 highest possibiity faces
bboxes = bboxes.astype(int)
bboxes = bboxes + [-1,-1,1,1] # personal choice
results, score, embs = face_recognize.infer(faces, targets)
for idx, bbox in enumerate(bboxes):
if False:
frame = draw_box_name(bbox, names[results[idx] + 1] + '_{:.2f}'.format(score[idx]), frame)
else:
frame = draw_box_name(bbox, names[results[idx] + 1], frame)
payload = cv2.imencode('.jpg', frame)[1].tobytes()
yield (b'--frame\r\n'
b'Content-Type: image/jpeg\r\n\r\n' + payload + b'\r\n')
ret, frame = vid_source.read()
def recognize_webcam(image_np):
# image = cv2.cvtColor(image_np, cv2.COLOR_BGR2RGB)
image = Image.fromarray(image_np)
# try:
bboxes, faces = face_recognize.align_multi(image)
# except:
# # pass
# bboxes = []
# faces = []
if len(bboxes) != 0:
bboxes = bboxes[:,:-1] #shape:[10,4],only keep 10 highest possibiity faces
bboxes = bboxes.astype(int)
bboxes = bboxes + [-1,-1,1,1] # personal choice
results, score, embs = face_recognize.infer(faces, targets)
for idx, bbox in enumerate(bboxes):
is_spoofing = face_recognize.is_spoofing(faces[idx])
status = "attack"
if not is_spoofing:
status = "genuine"
if len(que_name) > 15:
que_name.pop()
que_name.append(names[results[idx] + 1]+ '_{:.7s}'.format(status))
else:
que_name.append(names[results[idx] + 1]+ '_{:.7s}'.format(status))
if False:
image_np = draw_box_name(bbox, names[results[idx] + 1] + '_{:.2f}__{:.7s}'.format(score[idx], status), image_np)
else:
image_np = draw_box_name(bbox, names[results[idx] +1 ], image_np)
return image_np
def process_two_image(data):
folder_name = str(uuid.uuid1())
command = 'mkdir %s' % folder_name
subprocess.call(command, shell=True)
image_path_origin = download_file_by_url(data['image_url_origin'],
folder_name)
image_path_detection = download_file_by_url(data['image_url_detection'],
folder_name)
from api import face_recognize
conf = get_config()
face_recognize = face_recognize(conf)
face_recognize._raw_load_single_face(image_path_origin)
targets = face_recognize.embeddings
image = Image.open(image_path_detection)
submiter = [['image_url', 'x1', 'y1', 'x2', 'y2', 'result']]
try:
bboxes, faces = face_recognize.align_multi(image)
except:
bboxes = []
faces = []
if len(bboxes) > 0:
bboxes = bboxes[:, :-1]
bboxes = bboxes.astype(int)
bboxes = bboxes + [-1, -1, 1, 1]
results, score = face_recognize.infer(faces, targets)
for id, (re, sc) in enumerate(zip(results, score)):
if re != -1:
temp = {
'x1': bboxes[id][0],
'y1': bboxes[id][1],
'x2': bboxes[id][2],
'y2': bboxes[id][3],
'result': 1
}
temp = [
data['image_url_detection'], bboxes[id][0], bboxes[id][1],
bboxes[id][2], bboxes[id][3], 1
]
submiter.append(temp)
command = 'rm -rf %s' % folder_name
subprocess.call(command, shell=True)
df = pd.DataFrame.from_records(submiter)
headers = df.iloc[0]
df = pd.DataFrame(df.values[1:], columns=headers)
df = df.sort_values(by=['result'], ascending=False)
results = df.to_json(orient='records')
return results
def process_images(image_path='', path=''):
from api import face_recognize
conf = get_config()
face_recognize = face_recognize(conf)
face_recognize._raw_load_single_face(image_path)
targets = face_recognize.embeddings
submiter = [['image', 'x1', 'y1', 'x2', 'y2', 'result']]
list_file = glob.glob(path + '/*')
if os.path.isfile(list_file[0]) == False:
path = list_file[0]
print(path)
for img in tqdm(glob.glob(path + '/*')):
temp = [img.split('/')[-1], 0, 0, 0, 0, 0]
image = Image.open(img)
try:
bboxes, faces = face_recognize.align_multi(image)
except:
bboxes = []
faces = []
if len(bboxes) > 0:
bboxes = bboxes[:, :-1]
bboxes = bboxes.astype(int)
bboxes = bboxes + [-1, -1, 1, 1]
results, score = face_recognize.infer(faces, targets)
for id, (re, sc) in enumerate(zip(results, score)):
if re != -1:
temp = [
img.split('/')[-1], bboxes[id][0], bboxes[id][1],
bboxes[id][2], bboxes[id][3], 1
]
submiter.append(temp)
df = pd.DataFrame.from_records(submiter)
headers = df.iloc[0]
df = pd.DataFrame(df.values[1:], columns=headers)
df = df.sort_values(by=['result'], ascending=False)
results = df.to_json(orient='records')
return results
def recognize(image_path, targets, names):
frame = cv2.imread(image_path)
image = Image.fromarray(frame).convert('RGB')
try:
bboxes, faces = face_recognize.align_multi(image)
except:
bboxes = []
faces = []
result = {}
result['original'] = encode_image(image.copy())
if len(bboxes) != 0:
bboxes = bboxes[:,:-1] #shape:[10,4],only keep 10 highest possibiity faces
bboxes = bboxes.astype(int)
bboxes = bboxes + [-1,-1,1,1] # personal choice
results, score, embs = face_recognize.infer(faces, targets)
for idx, bbox in enumerate(bboxes):
result[names[results[idx] + 1]] = encode_image(faces[idx])
if False:
frame = draw_box_name(bbox, names[results[idx]+1] + '_{:.2f}'.format(score[idx]), frame)
else:
frame = draw_box_name(bbox, names[results[idx]+1], frame)
return result
isSuccess, frame = cap.read()
if isSuccess:
img_bg = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
image = Image.fromarray(img_bg)
try:
bboxes, faces = face_recognize.align_multi(image, thresholds = [0.5, 0.7, 0.8])
j+=1
except:
bboxes = []
faces = []
if len(bboxes) != 0:
bboxes = bboxes[:,:-1] #shape:[10,4],only keep 10 highest possibiity faces
bboxes = bboxes.astype(int)
bboxes = bboxes + [-1,-1,1,1] # personal choice
results, score, embs = face_recognize.infer(faces, targets)
for idx, bbox in enumerate(bboxes):
if results[idx] == -1 and args.save_unknow:
new_per = "%s/unknow_%s"%(conf.facebank_path, count_unknow)
if not os.path.exists(new_per):
os.mkdir(new_per)
faces[idx].save('%s/%s.jpg'%(new_per, datetime.now().date().strftime('%Y%m%d')))
targets = torch.cat((targets, embs[idx].unsqueeze(0)), dim=0)
names =np.append(names, 'unknow_%s'%count_unknow)
count_unknow+=1
if args.score:
frame = draw_box_name(bbox, names[results[idx]] + '_{:.2f}'.format(score[idx]), frame)
else: