def postprocess(result, img_meta, num_classes=81, rescale=True):
det_bboxes = result['boxes']
det_labels = result['labels']
det_masks = result.get('masks', None)
if rescale:
img_h, img_w = img_meta[0]['ori_shape'][:2]
scale = img_meta[0]['scale_factor']
det_bboxes[:, :4] /= scale
else:
img_h, img_w = img_meta[0]['img_shape'][:2]
det_bboxes[:, 0:4:2] = np.clip(det_bboxes[:, 0:4:2], 0, img_w - 1)
det_bboxes[:, 1:4:2] = np.clip(det_bboxes[:, 1:4:2], 0, img_h - 1)
bbox_results = bbox2result(det_bboxes, det_labels, num_classes)
if det_masks is not None:
segm_results = mask2result(
det_bboxes,
det_labels,
det_masks,
num_classes,
mask_thr_binary=0.5,
rle=True,
full_size=True,
img_size=(img_h, img_w))
return bbox_results, segm_results
return bbox_results
def postprocess(self,
det_bboxes,
det_labels,
det_masks,
img_meta,
rescale=False):
img_h, img_w = img_meta[0]['ori_shape'][:2]
num_classes = self.bbox_head.num_classes
scale_factor = img_meta[0]['scale_factor']
if isinstance(scale_factor, float):
scale_factor = np.asarray((scale_factor, ) * 4)
if rescale:
# Keep original image resolution unchanged
# and scale bboxes and masks to it.
if isinstance(det_bboxes, torch.Tensor):
scale_factor = det_bboxes.new_tensor(scale_factor)
det_bboxes[:, :4] /= scale_factor
else:
# Resize image to test resolution
# and keep bboxes and masks in test scale.
img_h = np.round(img_h * scale_factor[1]).astype(np.int32)
img_w = np.round(img_w * scale_factor[0]).astype(np.int32)
bbox_results = bbox2result(det_bboxes, det_labels, num_classes)
if self.with_mask:
segm_results = mask2result(
det_bboxes,
det_labels,
det_masks,
num_classes,
mask_thr_binary=self.test_cfg.rcnn.mask_thr_binary,
rle=True,
full_size=True,
img_size=(img_h, img_w))
return bbox_results, segm_results
return bbox_results