def __call__(self, gt_boxes, gt_labels):
if type(gt_boxes) is np.ndarray:
gt_boxes = torch.from_numpy(gt_boxes)
if type(gt_labels) is np.ndarray:
gt_labels = torch.from_numpy(gt_labels)
boxes, labels = box_utils.assign_priors(gt_boxes, gt_labels,
self.corner_form_priors, self.iou_threshold)
boxes = box_utils.corner_form_to_center_form(boxes)
locations = box_utils.convert_boxes_to_locations(boxes, self.center_form_priors, self.center_variance, self.size_variance)
return locations, labels
def __call__(self, gt_boxes, gt_labels):
if type(gt_boxes) is np.ndarray:
gt_boxes = torch.from_numpy(gt_boxes)
if type(gt_labels) is np.ndarray:
gt_labels = torch.from_numpy(gt_labels)
if (gt_boxes.shape[0] == 0):
raise ValueError("Error in ground truth box size: " +
str(gt_boxes.shape))
boxes, labels = box_utils.assign_priors(gt_boxes, gt_labels,
self.corner_form_priors,
self.iou_threshold)
boxes = box_utils.corner_form_to_center_form(boxes)
locations = box_utils.convert_boxes_to_locations(
boxes, self.center_form_priors, self.center_variance,
self.size_variance)
return locations, labels
def __call__(self, quad_gt,img_prior=None):
#将np array转为torch tensor
# if type(quad) is np.ndarray:
# quad = torch.from_numpy(quad)
priors_boxes,prior_boxed_labels = box_utils.assign_priors(quad_gt,self.quad_form_priors, self.iou_threshold,self.distance_threshold)
# print('prior_boxed_labels:',prior_boxed_labels)
true_default_box_index=np.where(prior_boxed_labels==1)
true_default_box=self.quad_form_priors[true_default_box_index] * 512
# print('self.quad_form_prior.shape:',np.shape(self.quad_form_priors))
tmp = self.quad_form_priors[382200:382270,:] * 512
# print('tmp.shape:',np.shape(tmp))
# print('true_default_box.shape:',np.shape(true_default_box))
if img_prior is not None:
for i in range(np.shape(true_default_box)[0]):
print(true_default_box[i])
# print(quad[i])
cv2.line(img_prior, (true_default_box[i][0], true_default_box[i][1]), (true_default_box[i][2], true_default_box[i][3]), (255, 0, 0), thickness=2)
cv2.line(img_prior, (true_default_box[i][2], true_default_box[i][3]), (true_default_box[i][4], true_default_box[i][5]), (255, 0, 0), thickness=2)
cv2.line(img_prior, (true_default_box[i][4], true_default_box[i][5]), (true_default_box[i][6], true_default_box[i][7]), (255, 0, 0), thickness=2)
cv2.line(img_prior, (true_default_box[i][6], true_default_box[i][7]), (true_default_box[i][0], true_default_box[i][1]), (255, 0, 0), thickness=2)
for i in range(np.shape(tmp)[0]):
# print('tmp[i]',tmp[i])
# print('i:',i)
cv2.line(img_prior, (tmp[i][0], tmp[i][1]), (tmp[i][2], tmp[i][3]), (0, 0, 255), thickness=2)
cv2.line(img_prior, (tmp[i][2], tmp[i][3]), (tmp[i][4], tmp[i][5]), (0, 0, 255), thickness=2)
cv2.line(img_prior, (tmp[i][4], tmp[i][5]), (tmp[i][6], tmp[i][7]), (0, 0, 255), thickness=2)
cv2.line(img_prior, (tmp[i][6], tmp[i][7]), (tmp[i][0], tmp[i][1]), (0, 0, 255), thickness=2)
cv2.imshow('img_prior', img_prior)
cv2.imwrite('test.jpg',img_prior)
cv2.waitKey()
# print('priors_boxes.shape',np.shape(priors_boxes))
locations = box_utils.convert_boxes_to_locations(priors_boxes, self.quad_form_priors, self.center_variance, self.size_variance)
#locations.size():[24564,8]
#prior_boxed_labels.size():[24564,]
return locations,prior_boxed_labels
