def _map_func(self,dp,is_training):
"""Data augmentation function."""
####customed here
try:
fname, ann = dp
image = cv2.imread(fname, cv2.IMREAD_COLOR)
if cfg.DATA.rgb:
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
label = np.array(ann)
if is_training:
image=self.random_crop_resize(image)
if random.uniform(0, 1) > 0.5:
image, _ = Mirror(image, label=None, symmetry=None)
if random.uniform(0, 1) > 0.5:
angle = random.uniform(-45, 45)
image, _ = Rotate_aug(image, label=None, angle=angle)
if random.uniform(0, 1) > 1.:
strength = random.uniform(0, 50)
image, _ = Affine_aug(image, strength=strength, label=None)
if random.uniform(0, 1) > 0.5:
image=self.color_augmentor(image)
if random.uniform(0, 1) > 1.0:
image=pixel_jitter(image,15)
if random.uniform(0, 1) > 0.5:
image = Img_dropout(image, 0.2)
else:
###centercrop
image = self.center_crop(image)
label = label.astype(np.int64)
image= image.astype(np.uint8)
except:
logger.info('some err happended with %s'%fname, ' but handled with -1')
image=np.zeros(shape=[cfg.MODEL.hin,cfg.MODEL.win,3],dtype=np.uint8)
label = np.array(-1,dtype=np.int64)
return image, label
def _map_func(self,dp,is_training):
"""Data augmentation function."""
####customed here
fname= dp['image_path']
keypoints=dp['keypoints']
bbox =dp['bbox']
# attr =dp['attr']
#### 300W
if keypoints is not None:
image = cv2.imread(fname, cv2.IMREAD_COLOR)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
label = np.array(keypoints, dtype=np.float).reshape((-1, 2))
bbox = np.array(bbox)
### random crop and resize
crop_image, label = self.augmentationCropImage(image, bbox, label, is_training)
if is_training:
if random.uniform(0, 1) > 0.5:
crop_image, label = Mirror(crop_image, label=label, symmetry=cfg.DATA.symmetry)
if random.uniform(0, 1) > 0.0:
angle = random.uniform(-45, 45)
crop_image, label = Rotate_aug(crop_image, label=label, angle=angle)
if random.uniform(0, 1) > 0.5:
strength = random.uniform(0, 50)
crop_image, label = Affine_aug(crop_image, strength=strength, label=label)
if random.uniform(0, 1) > 0.5:
crop_image=self.color_augmentor(crop_image)
if random.uniform(0, 1) > 0.5:
crop_image=pixel_jitter(crop_image,15)
if random.uniform(0, 1) > 0.5:
crop_image = Img_dropout(crop_image, 0.2)
if random.uniform(0, 1) > 0.5:
crop_image = Padding_aug(crop_image, 0.3)
#######head pose
reprojectdst, euler_angle = get_head_pose(label, crop_image)
PRY = euler_angle.reshape([-1]).astype(np.float32) / 90.
######cla_label
cla_label = np.zeros([4])
if np.sqrt(np.square(label[37, 0] - label[41, 0]) +
np.square(label[37, 1] - label[41, 1])) / cfg.MODEL.hin < self.eye_close_thres \
or np.sqrt(np.square(label[38, 0] - label[40, 0]) +
np.square(label[38, 1] - label[40, 1])) / cfg.MODEL.hin < self.eye_close_thres:
cla_label[0] = 1
if np.sqrt(np.square(label[43, 0] - label[47, 0]) +
np.square(label[43, 1] - label[47, 1])) / cfg.MODEL.hin < self.eye_close_thres \
or np.sqrt(np.square(label[44, 0] - label[46, 0]) +
np.square(label[44, 1] - label[46, 1])) / cfg.MODEL.hin < self.eye_close_thres:
cla_label[1] = 1
if np.sqrt(np.square(label[61, 0] - label[67, 0]) +
np.square(label[61, 1] - label[67, 1])) / cfg.MODEL.hin < self.mouth_close_thres \
or np.sqrt(np.square(label[62, 0] - label[66, 0]) +
np.square(label[62, 1] - label[66, 1])) / cfg.MODEL.hin < self.mouth_close_thres \
or np.sqrt(np.square(label[63, 0] - label[65, 0]) +
np.square(label[63, 1] - label[65, 1])) / cfg.MODEL.hin < self.mouth_close_thres:
cla_label[2] = 1
### mouth open big 1 mean true
if np.sqrt(np.square(label[62, 0] - label[66, 0]) +
np.square(label[62, 1] - label[66, 1])) / cfg.MODEL.hin > self.big_mouth_open_thres:
cla_label[3] = 1
crop_image_height, crop_image_width, _ = crop_image.shape
label = label.astype(np.float32)
label[:, 0] = label[:, 0] / crop_image_width
label[:, 1] = label[:, 1] / crop_image_height
crop_image = crop_image.astype(np.float32)
label = label.reshape([-1]).astype(np.float32)
cla_label = cla_label.astype(np.float32)
label = np.concatenate([label, PRY, cla_label], axis=0)
# else:
# image = cv2.imread(fname, cv2.IMREAD_COLOR)
# image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
#
# bbox = np.array(bbox)
#
# ### random crop and resize
# crop_image, _ = self.augmentationCropImage(image, bbox, np.array([[0,0]], dtype=np.float).reshape((-1, 2)), is_training)
#
# if is_training:
#
# if random.uniform(0, 1) > 0.5:
# crop_image, _ = Mirror(crop_image, label=None, symmetry=None)
# if random.uniform(0, 1) > 0.0:
# angle = random.uniform(-45, 45)
# crop_image, _ = Rotate_aug(crop_image, label=None, angle=angle)
#
# if random.uniform(0, 1) > 0.5:
# strength = random.uniform(0, 50)
# crop_image, _ = Affine_aug(crop_image, strength=strength, label=None)
#
# if random.uniform(0, 1) > 0.5:
# crop_image = self.color_augmentor(crop_image)
# if random.uniform(0, 1) > 0.5:
# crop_image = pixel_jitter(crop_image, 15)
# if random.uniform(0, 1) > 0.5:
# crop_image = Img_dropout(crop_image, 0.2)
#
# if random.uniform(0, 1) > 0.5:
# crop_image = Padding_aug(crop_image, 0.3)
#
# #######head pose
#
# crop_image = crop_image.astype(np.float32)
#
# label=np.zeros(shape=cfg.MODEL.out_channel-2)-1
# label[-4:]=np.array(attr)
#
# ### index 0 means things come from keypoints, 1 is from celeba
# weights = np.array([0, 1])
#
# label = np.concatenate([label, weights], axis=0)
return crop_image, label
def _map_func(self, dp, is_training):
"""Data augmentation function."""
####customed here
fname, ann = dp
image = cv2.imread(fname, cv2.IMREAD_COLOR)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
label = np.array([ann.split(' ')], dtype=np.float).reshape((-1, 2))
bbox = np.array([
np.min(label[:, 0]),
np.min(label[:, 1]),
np.max(label[:, 0]),
np.max(label[:, 1])
])
### random crop and resize
crop_image, label = self.augmentationCropImage(image, bbox, label,
is_training)
if is_training:
if random.uniform(0, 1) > 0.5:
crop_image, label = Mirror(crop_image,
label=label,
symmetry=cfg.DATA.symmetry)
if random.uniform(0, 1) > 0.0:
angle = random.uniform(-45, 45)
crop_image, label = Rotate_aug(crop_image,
label=label,
angle=angle)
if random.uniform(0, 1) > 0.5:
strength = random.uniform(0, 50)
crop_image, label = Affine_aug(crop_image,
strength=strength,
label=label)
if random.uniform(0, 1) > 0.5:
crop_image = self.color_augmentor(crop_image)
if random.uniform(0, 1) > 0.5:
crop_image = pixel_jitter(crop_image, 15)
if random.uniform(0, 1) > 0.5:
crop_image = Img_dropout(crop_image, 0.2)
if random.uniform(0, 1) > 0.5:
crop_image = Padding_aug(crop_image, 0.3)
#######head pose
reprojectdst, euler_angle = get_head_pose(label, crop_image)
PRY = euler_angle.reshape([-1]).astype(np.float32) / 90.
######cla_label
cla_label = np.zeros([4])
if np.sqrt(np.square(label[37, 0] - label[41, 0]) +
np.square(label[37, 1] - label[41, 1])) / cfg.MODEL.hin < self.eye_close_thres \
or np.sqrt(np.square(label[38, 0] - label[40, 0]) +
np.square(label[38, 1] - label[40, 1])) / cfg.MODEL.hin < self.eye_close_thres:
cla_label[0] = 1
if np.sqrt(np.square(label[43, 0] - label[47, 0]) +
np.square(label[43, 1] - label[47, 1])) / cfg.MODEL.hin < self.eye_close_thres \
or np.sqrt(np.square(label[44, 0] - label[46, 0]) +
np.square(label[44, 1] - label[46, 1])) / cfg.MODEL.hin < self.eye_close_thres:
cla_label[1] = 1
if np.sqrt(np.square(label[61, 0] - label[67, 0]) +
np.square(label[61, 1] - label[67, 1])) / cfg.MODEL.hin < self.mouth_close_thres \
or np.sqrt(np.square(label[62, 0] - label[66, 0]) +
np.square(label[62, 1] - label[66, 1])) / cfg.MODEL.hin < self.mouth_close_thres \
or np.sqrt(np.square(label[63, 0] - label[65, 0]) +
np.square(label[63, 1] - label[65, 1])) / cfg.MODEL.hin < self.mouth_close_thres:
cla_label[2] = 1
### mouth open big 1 mean true
if np.sqrt(
np.square(label[62, 0] - label[66, 0]) +
np.square(label[62, 1] - label[66, 1])
) / cfg.MODEL.hin > self.big_mouth_open_thres:
cla_label[3] = 1
crop_image_height, crop_image_width, _ = crop_image.shape
label = label.astype(np.float32)
label[:, 0] = label[:, 0] / crop_image_width
label[:, 1] = label[:, 1] / crop_image_height
crop_image = crop_image.astype(np.float32)
label = label.reshape([-1]).astype(np.float32)
cla_label = cla_label.astype(np.float32)
label = np.concatenate([label, PRY, cla_label], axis=0)
return crop_image, label
def _map_func(self,dp,is_training):
"""Data augmentation function."""
####customed here
cur_data_info=dp.rstrip().split(' ')
fname= cur_data_info[0]
label=cur_data_info[1:4]
label=[int(x) for x in label]
# attr =dp['attr']
#### 300W
image = cv2.imread(fname, cv2.IMREAD_COLOR)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
label = np.array(label, dtype=np.float)
if is_training:
if random.uniform(0, 1) > 0.5:
image=Random_crop(image,shrink=0.2)
# if random.uniform(0, 1) > 0.5.:
# image, _ = Mirror(image, label=None, symmetry=None)
if random.uniform(0, 1) > 0.0:
angle = random.uniform(-30, 30)
image, _ = Rotate_aug(image, label=None, angle=angle)
if random.uniform(0, 1) > 0.5:
strength = random.uniform(0, 50)
image, _ = Affine_aug(image, strength=strength, label=None)
if random.uniform(0, 1) > 0.5:
image=self.color_augmentor(image)
if random.uniform(0, 1) > 0.5:
image=pixel_jitter(image,15)
if random.uniform(0, 1) > 0.5:
image = Img_dropout(image, 0.2)
if random.uniform(0, 1) > 0.5:
image = Padding_aug(image, 0.3)
interp_methods = [cv2.INTER_LINEAR, cv2.INTER_CUBIC, cv2.INTER_AREA, cv2.INTER_NEAREST,
cv2.INTER_LANCZOS4]
interp_method = random.choice(interp_methods)
image = cv2.resize(image, (cfg.MODEL.win, cfg.MODEL.hin), interpolation=interp_method)
#######head pose
if cfg.MODEL.channel==1:
image = image.astype(np.uint8)
image=cv2.cvtColor(image,cv2.COLOR_RGB2GRAY)
image=np.expand_dims(image,-1)
image = np.transpose(image,axes=[2,0,1])
label = label.reshape([-1]).astype(np.float32)
image = image.astype(np.float32)
if not cfg.TRAIN.vis :
image=image/255.
return image, label
def _map_func(self, dp, is_training):
"""Data augmentation function."""
####customed here
fname = dp['image_path']
keypoints = dp['keypoints']
# attr =dp['attr']
#### 300W
if keypoints is not None:
image = cv2.imread(fname, cv2.IMREAD_COLOR)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
label = np.array(keypoints, dtype=np.float).reshape((-1, 2))
#bbox = np.array(bbox)
### random crop and resize
#crop_image, label = self.augmentationCropImage(image, bbox, label, is_training)
crop_image, label = self.preprocess(image, cfg.MODEL.hin,
cfg.MODEL.win, label)
if is_training:
if random.uniform(0, 1) > 0.5:
crop_image, label = Mirror(crop_image,
label=label,
symmetry=cfg.DATA.symmetry)
if random.uniform(0, 1) > 0.0:
angle = random.uniform(-45, 45)
crop_image, label = Rotate_aug(crop_image,
label=label,
angle=angle)
if random.uniform(0, 1) > 0.5:
strength = random.uniform(0, 50)
crop_image, label = Affine_aug(crop_image,
strength=strength,
label=label)
if random.uniform(0, 1) > 0.5:
crop_image = self.color_augmentor(crop_image)
if random.uniform(0, 1) > 0.5:
crop_image = pixel_jitter(crop_image, 15)
if random.uniform(0, 1) > 0.5:
crop_image = Img_dropout(crop_image, 0.2)
if random.uniform(0, 1) > 0.5:
crop_image = Padding_aug(crop_image, 0.3)
look = dp['look_vec']
crop_image_height, crop_image_width, _ = crop_image.shape
label = label.astype(np.float32)
label[:, 0] = label[:, 0] / crop_image_width
label[:, 1] = label[:, 1] / crop_image_height
crop_image = crop_image.astype(np.float32)
label = label.reshape([-1]).astype(np.float32)
##cla_label = cla_label.astype(np.float32)
look = np.array(look, dtype=np.float32)
label = np.concatenate([label, look], axis=0)
# else:
# image = cv2.imread(fname, cv2.IMREAD_COLOR)
# image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
#
# bbox = np.array(bbox)
#
# ### random crop and resize
# crop_image, _ = self.augmentationCropImage(image, bbox, np.array([[0,0]], dtype=np.float).reshape((-1, 2)), is_training)
#
# if is_training:
#
# if random.uniform(0, 1) > 0.5:
# crop_image, _ = Mirror(crop_image, label=None, symmetry=None)
# if random.uniform(0, 1) > 0.0:
# angle = random.uniform(-45, 45)
# crop_image, _ = Rotate_aug(crop_image, label=None, angle=angle)
#
# if random.uniform(0, 1) > 0.5:
# strength = random.uniform(0, 50)
# crop_image, _ = Affine_aug(crop_image, strength=strength, label=None)
#
# if random.uniform(0, 1) > 0.5:
# crop_image = self.color_augmentor(crop_image)
# if random.uniform(0, 1) > 0.5:
# crop_image = pixel_jitter(crop_image, 15)
# if random.uniform(0, 1) > 0.5:
# crop_image = Img_dropout(crop_image, 0.2)
#
# if random.uniform(0, 1) > 0.5:
# crop_image = Padding_aug(crop_image, 0.3)
#
# #######head pose
#
# crop_image = crop_image.astype(np.float32)
#
# label=np.zeros(shape=cfg.MODEL.out_channel-2)-1
# label[-4:]=np.array(attr)
#
# ### index 0 means things come from keypoints, 1 is from celeba
# weights = np.array([0, 1])
#
# label = np.concatenate([label, weights], axis=0)
return crop_image, label
def _map_func(self, dp, is_training):
"""Data augmentation function."""
####customed here
try:
fname, annos = dp
image = cv2.imread(fname, cv2.IMREAD_COLOR)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
labels = annos.split(' ')
boxes = []
for label in labels:
bbox = np.array(label.split(','), dtype=np.float)
boxes.append([bbox[0], bbox[1], bbox[2], bbox[3], bbox[4]])
boxes = np.array(boxes, dtype=np.float)
if is_training:
sample_dice = random.uniform(0, 1)
if sample_dice > 0.8 and sample_dice <= 1:
image, boxes = Random_scale_withbbox(
image,
boxes,
target_shape=[cfg.DATA.hin, cfg.DATA.win],
jitter=0.5)
elif sample_dice > 0.4 and sample_dice <= 0.8:
boxes_ = boxes[:, 0:4]
klass_ = boxes[:, 4:]
image, boxes_, klass_ = dsfd_aug(image, boxes_, klass_)
image = image.astype(np.uint8)
boxes = np.concatenate([boxes_, klass_], axis=1)
else:
boxes_ = boxes[:, 0:4]
klass_ = boxes[:, 4:]
image, boxes_, klass_ = baidu_aug(image, boxes_, klass_)
image = image.astype(np.uint8)
boxes = np.concatenate([boxes_, klass_], axis=1)
if random.uniform(0, 1) > 0.5:
image, boxes = Random_flip(image, boxes)
if random.uniform(0, 1) > 0.5:
boxes_ = boxes[:, 0:4]
klass_ = boxes[:, 4:]
angel = random.uniform(-30, 30)
image, boxes_ = Rotate_with_box(image, angel, boxes_)
boxes = np.concatenate([boxes_, klass_], axis=1)
if random.uniform(0, 1) > 0.5:
image = self.color_augmentor(image)
if random.uniform(0, 1) > 0.5:
image = pixel_jitter(image, 15)
else:
boxes_ = boxes[:, 0:4]
klass_ = boxes[:, 4:]
image, shift_x, shift_y = Fill_img(image,
target_width=cfg.DATA.win,
target_height=cfg.DATA.hin)
boxes_[:, 0:4] = boxes_[:, 0:4] + np.array(
[shift_x, shift_y, shift_x, shift_y], dtype='float32')
h, w, _ = image.shape
boxes_[:, 0] /= w
boxes_[:, 1] /= h
boxes_[:, 2] /= w
boxes_[:, 3] /= h
image = image.astype(np.uint8)
image = cv2.resize(image, (cfg.DATA.win, cfg.DATA.hin))
boxes_[:, 0] *= cfg.DATA.win
boxes_[:, 1] *= cfg.DATA.hin
boxes_[:, 2] *= cfg.DATA.win
boxes_[:, 3] *= cfg.DATA.hin
image = image.astype(np.uint8)
boxes = np.concatenate([boxes_, klass_], axis=1)
if boxes.shape[0] == 0 or np.sum(image) == 0:
boxes_ = np.array([[0, 0, 100, 100]])
klass_ = np.array([0])
else:
boxes_ = np.array(boxes[:, 0:4], dtype=np.float32)
klass_ = np.array(boxes[:, 4], dtype=np.int64)
except:
logger.warn('there is an err with %s' % fname)
traceback.print_exc()
image = np.zeros(shape=(cfg.DATA.hin, cfg.DATA.win, 3),
dtype=np.float32)
boxes_ = np.array([[0, 0, 100, 100]])
klass_ = np.array([0])
return image, boxes_, klass_