def load_noonan_val_pair(path, rootdir, transform, image_size=[112,112]):
# path: path for original images, str
# rootdir: path to store images and issame_list, Path
mtcnn = MTCNN()
print('mtcnn loaded')
if not rootdir.exists():
rootdir.mkdir()
images = os.listdir(path)
comb_size = len(images) * (len(images) - 1) / 2
pairs = combinations(images, 2)
data = bcolz.fill([int(comb_size * 2), 3, image_size[0], image_size[1]], dtype=np.float32, rootdir=rootdir, mode='w')
issame_list = np.zeros(int(comb_size))
i = 0
for pair in pairs:
img0 = Image.open(os.path.join(path, pair[0]))
if img0.size != (112, 112):
img0 = mtcnn.align(img0)
img1 = Image.open(os.path.join(path, pair[1]))
if img1.size != (112, 112):
img1 = mtcnn.align(img1)
data[2*i, ...] = transform(img0)
data[2*i + 1, ...] = transform(img1)
if ('noonan' in pair[0] and 'noonan' in pair[1]) or ('normal' in pair[0] and 'normal' in pair[1]):
issame_list[i] = 1
i += 1
if i % 1000 == 0:
print('loading noonan', i)
print(data.shape)
np.save(str(rootdir)+'_list', np.array(issame_list))
return data, issame_list
def alignMain(args):
LOG_DIR = '../Dataset/LogError/'
os.makedirs(os.path.dirname(LOG_DIR) + "/", exist_ok=True)
f = open(LOG_DIR+'AlignErrortxt', 'w')
if not os.path.exists(args.rawdataDir):
raise UserWarning("Input Dataset Directory does not exist!")
os.makedirs(os.path.dirname(args.outputDir) + "/", exist_ok=True)
dataset = load_dataset(args.rawdataDir)
device = 'cuda:0' if torch.cuda.is_available() else 'cpu'
detector = MTCNN(image_size=args.size, device=device)
for i in range(len(dataset)):
print("Working in folder " + dataset[i].name +" with index " + str(i), end='')
for path in dataset[i].paths:
_, file_name = os.path.split(path)
imgBGR = cv2.imread(path)
thumbnails = detector.align(img=imgBGR, select_largest=True)
if thumbnails is not None:
thumbnail = thumbnails[-1]
# thumbnail = cv2.cvtColor(np.float32(thumbnail), cv2.COLOR_RGB2BGR)
OUTPUT_DIR = args.outputDir + dataset[i].name + "/"
os.makedirs(os.path.dirname(OUTPUT_DIR) + "/", exist_ok=True)
str1 = "{}{}".format(OUTPUT_DIR, file_name)
cv2.imwrite(str1, thumbnail)
else:
f.write("Unable to detect faces: {}/{}. \n".format(dataset[i].name, file_name))
print(' ... Done!')
f.close()
def get_sg_faces(source_path, save_path):
# st: stylegan
# os.makedirs(save_path, exist_ok=True)
fake_images = [item for item in os.listdir(source_path)]
mtcnn = MTCNN()
for img in fake_images:
bboxes, _ = mtcnn.align_multi(Image.open(source_path + os.sep + img), 1, 30)
# the last two params: conf.face_limit, conf.min_face_size
if bboxes.shape[0] != 1:
print(img + ':', bboxes.shape)
else:
new_img = mtcnn.align(Image.open(source_path + os.sep + img), (112,112))
subfolder = os.path.basename(img).split('_')[0]
os.makedirs(save_path + os.sep + subfolder, exist_ok=True)
new_img.save(os.path.join(save_path, subfolder, img))
def register(user_id):
data_path = Path('data')
save_path = data_path / 'facebank' / user_id
fetch_path = data_path / 'dataset' / user_id
images = load_images_from_folder(fetch_path)
print(images)
if not save_path.exists():
save_path.mkdir()
mtcnn = MTCNN()
count = 0
face_id = user_id
count = 0
for img in images:
frame = img
p = Image.fromarray(frame[..., ::-1])
try:
warped_face = np.array(mtcnn.align(p))[..., ::-1]
cv2.imwrite(
"data/facebank/" + str(face_id) + '/' + str(face_id) + '_' +
str(count) + ".jpg", warped_face)
count += 1
#cv2.imwrite(str(save_path/'{}.jpg'.format(str(datetime.now())[:-7].replace(":","-").replace(" ","-"))), warped_face)
except:
result = {
"_result": "Error",
"_message": "Unable to detect the face"
}
if count == len(images):
result = {
"_result": "success",
"_message": "User Registered Successfully"
}
conf = get_config(False)
learner = face_learner(conf, True)
learner.load_state(conf, 'cpu_final.pth', True, True)
learner.model.eval()
#print('learner loaded')
targets, names = prepare_facebank(conf, learner.model, mtcnn, user_id)
#print('facebank updated')
return result
def get_train_dataset_gan(imgs_folder, target_folder, target_size):
# if 'divided' in imgs_folder:
for sub_folder in os.listdir(imgs_folder):
os.makedirs(target_folder + os.sep + sub_folder, exist_ok=True)
# else:
# os.makedirs(target_folder, exist_ok=True)
imgs=[]
for root, _, fnames in sorted(os.walk(imgs_folder)):
for fname in fnames:
if is_image_file(fname):
path = os.path.join(root, fname)
imgs.append((path, fname))
mtcnn = MTCNN()
for img in imgs:
bboxes, _ = mtcnn.align_multi(Image.open(img[0]), 1, 30)
# the last two params: conf.face_limit, conf.min_face_size
if bboxes.shape[0] > 0:
new_img = mtcnn.align(Image.open(img[0]), target_size)
# if 'divided' in imgs_folder:
new_img.save(target_folder + os.sep + '/'.join(img[0].strip().split(os.sep)[-2:]))
def align_images(base_dir, output_dir):
if not os.path.exists(output_dir):
os.mkdir(output_dir)
mtcnn = MTCNN()
patients_list = os.listdir(base_dir)
for ppl in patients_list:
patient_folder = os.path.join(base_dir, ppl)
new_folder = os.path.join(output_dir, ppl)
if not os.path.exists(new_folder):
os.mkdir(new_folder)
print("Creating new folder:", new_folder)
img_list = os.listdir(patient_folder)
for img in img_list:
image = Image.open(os.path.join(patient_folder,
img)).convert('RGB')
if image.size != (112, 112):
new_image = mtcnn.align(image)
# new_image = image.resize((112, 112))
new_image.save(os.path.join(new_folder, img))
print(new_image.size)
print('Saving', img)
if isSuccess:
frame_text = cv2.putText(frame,
'Press t to take a picture,q to quit.....',
(10,100),
cv2.FONT_HERSHEY_SIMPLEX,
2,
(0,255,0),
3,
cv2.LINE_AA)
cv2.imshow("My Capture",frame_text)
# 实现按下“t”键拍照
if cv2.waitKey(1):
p = Image.fromarray(frame[...,::-1])
try:
warped_face = np.array(mtcnn.align(p))[...,::-1]
bboxes, faces = mtcnn.align_multi(p, conf.face_limit, conf.min_face_size)
print('-----------box count {}', len(bboxes))
bboxes = bboxes[:,:-1] #shape:[10,4],only keep 10 highest possibiity faces
bboxes = bboxes.astype(int)
bboxes = bboxes + [-1,-1,1,1] # personal choice
for box in bboxes:
frame_text = cv2.rectangle(frame_text,(box[0],box[1]),(box[2],box[3]),(0,0,255),6)
cv2.imwrite(str(save_path/'{}.jpg'.format(str(datetime.now())[:-7].replace(":","-").replace(" ","-"))), warped_face)
except:
print('no face captured')
if cv2.waitKey(1)&0xFF == ord('q'):
break
mtcnn = MTCNN()
print('mtcnn loaded')
# inital camera
img = []
img.append(cv2.imread('data/input/evans/evans_p.jpg'))
img.append(cv2.imread('data/input/hermsworth/hermsworth_p.jpg'))
img.append(cv2.imread('data/input/jeremy/jeremy.jpg'))
img.append(cv2.imread('data/input/mark/mark.jpg'))
img.append(cv2.imread('data/input/olsen/olsen.jpg'))
faces = []
re_img = Image.fromarray(img[0][..., ::-1])
re_img = mtcnn.align(re_img)
tolist_face = np.array(re_img).tolist()
URL = server + "register"
json_feed = {'face_image': tolist_face}
response = requests.post(URL, json=json_feed)
print(response)
tolist_face = img[1].tolist()
URL = server + "getframe"
json_feed = {'face_list': tolist_face}
response = requests.post(URL, json=json_feed)
'''
for i in range(5):
image = Image.fromarray(img[i][...,::-1]) #bgr to rgb
# image = Image.fromarray(img[i])
print('facebank updated')
else:
targets, ftoid, idinfo = load_facebank(conf)
print('facebank loaded')
faces = []
predfns = []
with open(args.file) as f:
imgfiles = list(map(str.strip, f.readlines()))
for imgfn in imgfiles:
try:
face = Image.open(imgfn)
except:
print('cannot open query image file {}'.format(imgfn))
continue
try:
face = mtcnn.align(face)
except:
print('mtcnn failed for {}'.format(imgfn))
face = face.resize((112, 112), Image.ANTIALIAS)
#data = np.array((cv2.cvtColor(np.asarray(face), cv2.COLOR_RGB2GRAY),)*3).T
#face = Image.fromarray(data)
data = np.array(face)
face = Image.fromarray(data[:, :, ::-1])
faces.append(face)
results, score, d = learner.infer(conf, faces, targets, args.tta)
print(score)
for idx, imgfn in enumerate(imgfiles):
i = results[idx]
print("For {} found face {}".format(
imgfn, "Unknown" if i == -1 else idinfo[i][1]))
print(d[idx], d[idx][i])
IMAGE_PATH1 = '../Dataset/Test/3.jpg'
device = 'cuda:0' if torch.cuda.is_available() else 'cpu'
detector = MTCNN(image_size=160, device=device)
model = MobileNetV2(128, alpha=1.0)
model.load_weights(TRAINED_MODEL).expect_partial()
with open(SVM_MODEL, 'rb') as infile:
(svm_model, class_names) = pickle.load(infile)
img = cv2.imread(IMAGE_PATH1)
if max(img.shape[0], img.shape[1]) > 900:
scale_percent = 900 / max(img.shape[0], img.shape[1])
img = img_resize(img, scale_percent)
#faces = detector.align(img=img, select_largest=False, save_path='../../Test/')
faces = detector.align(img=img, select_largest=False)
if faces is not None:
for face in faces:
face = np.float32(face)
face_embedding = model(face, True)
predictions = svm_model.predict_proba(face_embedding)
best_class_idxs = np.argmax(predictions, axis=1)
best_class_probabilities = predictions[np.arange(len(best_class_idxs)),
best_class_idxs]
for i in range(len(best_class_idxs)):
print('%4d %s: %.3f' % (i, class_names[best_class_idxs[i]],
best_class_probabilities[i]))
img1 = cv2.imread(IMAGE_PATH1)
if max(img1.shape[0], img1.shape[1]) > 900:
scale_percent = 900 / max(img1.shape[0], img1.shape[1])
img1 = img_resize(img1, scale_percent)
img2 = cv2.imread(IMAGE_PATH2)
if max(img2.shape[0], img2.shape[1]) > 900:
scale_percent = 900 / max(img2.shape[0], img2.shape[1])
img2 = img_resize(img2, scale_percent)
device = 'cuda:0' if torch.cuda.is_available() else 'cpu'
detector = MTCNN(image_size=160, device=device)
model = MobileNetV2(128, alpha=1.0)
model.load_weights(TRAINED_MODEL)
faces1 = detector.align(img=img1, select_largest=True)
faces2 = detector.align(img=img2, select_largest=True)
if faces1 is not None:
for face1 in faces1:
face1 = np.float32(face1)
face_embedding1 = model(face1, True)
print(face_embedding1)
if faces2 is not None:
for face2 in faces2:
face2 = np.float32(face2)
face_embedding2 = model(face2, True)
print(face_embedding2)