def pull_item(self, index):
"""
Returns image at index in torch tensor form (RGB),
corresponding normalized annotation in 2d array [[xmin, ymin, xmax, ymax, label_ind],
... ],
height and width of image
"""
img_path = list(self.annotation.keys())[index]
img = cv2.imread(img_path, cv2.IMREAD_COLOR)
assert img is not None
height, width, _ = img.shape
boxes = np.asarray(
[anno['bbox'] for anno in self.annotation[img_path]])
labels = np.asarray(
[anno['label_idx'] for anno in self.annotation[img_path]])
if self.scale_bboxes:
boxes /= np.array([width, height, width, height])
if not boxes.size:
logger.error("error: no annotation on image")
sys.exit(-1)
if self.target_transform is not None:
annotation = self.target_transform(self.annotation, width, height)
if self.transform is not None:
img, boxes, labels = self.transform(img, boxes, labels)
if self.rgb:
img = img[:, :, (2, 1, 0)]
annotation = np.hstack((boxes, np.expand_dims(labels, axis=1)))
return torch.from_numpy(img).permute(2, 0,
1), annotation, height, width
def load_weights(self, base_file):
_, ext = os.path.splitext(base_file)
if ext == '.pkl' or '.pth':
logger.debug('Loading weights into state dict...')
self.load_state_dict(torch.load(base_file,
map_location=lambda storage, loc: storage))
logger.debug('Finished!')
else:
logger.error('Sorry only .pth and .pkl files supported.')
def load_weights(self, base_file):
_, ext = os.path.splitext(base_file)
if ext == '.pkl' or '.pth':
logger.debug('Loading weights into state dict...')
#
# ** WARNING: torch.load functionality uses Python's pickling facilities that
# may be used to perform arbitrary code execution during unpickling. Only load the data you
# trust.
#
self.load_state_dict(
torch.load(base_file,
map_location=lambda storage, loc: storage))
logger.debug('Finished!')
else:
logger.error('Sorry only .pth and .pkl files supported.')