def aug_image(self, train_instance, jitter):
image_name = train_instance['filename']
if self._config['IMAGE_C'] == 1:
image = cv2.imread(image_name, cv2.IMREAD_GRAYSCALE)
elif self._config['IMAGE_C'] == 3:
image = cv2.imread(image_name)
else:
raise ValueError("Invalid number of image channels.")
if image is None:
print('Cannot find ', image_name)
if self._callback is not None:
image, train_instance = self._callback(image, train_instance)
h = image.shape[0]
w = image.shape[1]
all_objs = copy.deepcopy(train_instance['object'])
if jitter:
bbs = []
for i, obj in enumerate(all_objs):
xmin = obj['xmin']
ymin = obj['ymin']
xmax = obj['xmax']
ymax = obj['ymax']
# use label field to later match it with final boxes
bbs.append(BoundingBox(x1=xmin, x2=xmax, y1=ymin, y2=ymax, label=i))
bbs = BoundingBoxesOnImage(bbs, shape=image.shape)
image, bbs = self._aug_pipe(image=image, bounding_boxes=bbs)
bbs = bbs.remove_out_of_image().clip_out_of_image()
if len(bbs) < len(all_objs):
print("Some boxes were removed during augmentations.")
filtered_objs = []
for bb in bbs.bounding_boxes:
obj = all_objs[bb.label]
obj['xmin'] = bb.x1
obj['xmax'] = bb.x2
obj['ymin'] = bb.y1
obj['ymax'] = bb.y2
filtered_objs.append(obj)
all_objs = filtered_objs
# resize the image to standard size
image = cv2.resize(image, (self._config['IMAGE_W'], self._config['IMAGE_H']))
if self._config['IMAGE_C'] == 1:
image = image[..., np.newaxis]
image = image[..., ::-1] # make it RGB (it is important for normalization of some backends)
# fix object's position and size
for obj in all_objs:
for attr in ['xmin', 'xmax']:
obj[attr] = int(obj[attr] * float(self._config['IMAGE_W']) / w)
obj[attr] = max(min(obj[attr], self._config['IMAGE_W']), 0)
for attr in ['ymin', 'ymax']:
obj[attr] = int(obj[attr] * float(self._config['IMAGE_H']) / h)
obj[attr] = max(min(obj[attr], self._config['IMAGE_H']), 0)
return image, all_objs
def aug_image(self, train_instance, jitter):
image_name = train_instance['filename']
if self._config['IMAGE_C'] == 1:
image = cv2.imread(image_name, cv2.IMREAD_GRAYSCALE)
elif self._config['IMAGE_C'] == 3:
image = cv2.imread(image_name)
else:
raise ValueError("Invalid number of image channels.")
if image is None:
print('Cannot find ', image_name)
if self._callback is not None:
image, train_instance = self._callback(image, train_instance)
h = image.shape[0]
w = image.shape[1]
all_objs = copy.deepcopy(train_instance['object'])
if jitter:
bbs = []
for obj in all_objs:
xmin = obj['xmin']
ymin = obj['ymin']
xmax = obj['xmax']
ymax = obj['ymax']
bbs.append(BoundingBox(x1=xmin, x2=xmax, y1=ymin, y2=ymax))
bbs = BoundingBoxesOnImage(bbs, shape=image.shape)
image, bbs = self._aug_pipe(image=image, bounding_boxes=bbs)
bbs = bbs.remove_out_of_image().clip_out_of_image()
if len(all_objs) != 0:
for i in range(len(bbs.bounding_boxes)):
all_objs[i]['xmin'] = bbs.bounding_boxes[i].x1
all_objs[i]['xmax'] = bbs.bounding_boxes[i].x2
all_objs[i]['ymin'] = bbs.bounding_boxes[i].y1
all_objs[i]['ymax'] = bbs.bounding_boxes[i].y2
# resize the image to standard size
image = cv2.resize(image,
(self._config['IMAGE_W'], self._config['IMAGE_H']))
if self._config['IMAGE_C'] == 1:
image = image[..., np.newaxis]
image = image[..., ::-1]
# fix object's position and size
for obj in all_objs:
for attr in ['xmin', 'xmax']:
obj[attr] = int(obj[attr] * float(self._config['IMAGE_W']) / w)
obj[attr] = max(min(obj[attr], self._config['IMAGE_W']), 0)
for attr in ['ymin', 'ymax']:
obj[attr] = int(obj[attr] * float(self._config['IMAGE_H']) / h)
obj[attr] = max(min(obj[attr], self._config['IMAGE_H']), 0)
return image, all_objs
def aug_image(self, image, objs, jitter=True):
h = image.shape[0]
w = image.shape[1]
all_objs = copy.deepcopy(objs)
if jitter:
bbs = []
for obj in all_objs:
# see parse image for the correspondence between numeric index and values
xmin = obj[0]
xmax = obj[1]
ymin = obj[2]
ymax = obj[3]
bbs.append(BoundingBox(x1=xmin, x2=xmax, y1=ymin, y2=ymax))
bbs = BoundingBoxesOnImage(bbs, shape=image.shape)
image, bbs = self._aug_pipe(image=image, bounding_boxes=bbs)
bbs = bbs.remove_out_of_image().clip_out_of_image()
if len(all_objs) != 0:
for i in range(len(bbs.bounding_boxes)):
all_objs[i][0] = bbs.bounding_boxes[i].x1 # xmin
all_objs[i][1] = bbs.bounding_boxes[i].x2 # xmax
all_objs[i][2] = bbs.bounding_boxes[i].y1 # ymin
all_objs[i][3] = bbs.bounding_boxes[i].y2 # ymax
# resize the image to standard size
image = cv2.resize(image, (self._config['IMAGE_W'], self._config['IMAGE_H']))
'''
if self._config['IMAGE_C'] == 1: #self._config['IMAGE_C']
image = image[..., np.newaxis]
'''
#image = image[..., ::-1]
# fix object's position and size
for obj in all_objs:
for attr in [0, 1]: #xmin, xmax
obj[attr] = int(obj[attr] * float(self._config['IMAGE_W']) / w)
obj[attr] = max(min(obj[attr], self._config['IMAGE_W']), 0)
for attr in [2, 3]: #ymin, ymax
obj[attr] = int(obj[attr] * float(self._config['IMAGE_H']) / h)
obj[attr] = max(min(obj[attr], self._config['IMAGE_H']), 0)
return image, all_objs