resolve conflict and move bbbox2results to one_stage

mmdet/models/dense_heads/anchor_head.py

@@ -736,6 +736,10 @@ def aug_test(self, feats, img_metas, rescale=False):
  Defaults to False.
 
  Returns:
- list[ndarray]: bbox results of each class
+ list[tuple[Tensor, Tensor]]: Each item in result_list is 2-tuple.
+ The first item is bboxes with shape (n, 5), where 5 represent
+ (tl_x, tl_y, br_x, br_y, score).
+ The shape of the second tensor in the tuple is labels
+ with shape (n,), Length of list should always be 1.
  """
  return self.aug_test_bboxes(feats, img_metas, rescale=rescale)

mmdet/models/dense_heads/base_dense_head.py

@@ -57,3 +57,20 @@ def forward_train(self,
  else:
  proposal_list = self.get_bboxes(*outs, img_metas, cfg=proposal_cfg)
  return losses, proposal_list
+
+ def simple_test(self, feats, img_metas, rescale=False):
+ """Test function without test-time augmentation.
+
+ Args:
+ feats (tuple[torch.Tensor]): Multi-level features from the
+ upstream network, each is a 4D-tensor.
+ img_metas (list[dict]): List of image information.
+ rescale (bool, optional): Whether to rescale the results.
+ Defaults to False.
+
+ Returns:
+ list[list[np.ndarray]]: BBox results of each image and classes.
+ The outer list corresponds to each image. The inner list
+ corresponds to each class.
+ """
+ return self.simple_test_bboxes(feats, img_metas, rescale=rescale)

mmdet/models/dense_heads/cascade_rpn_head.py

@@ -782,4 +782,5 @@ def simple_test_rpn(self, x, img_metas):
 
  def aug_test_rpn(self, x, img_metas):
  """Augmented forward test function."""
- raise NotImplementedError
+ raise NotImplementedError(
+ 'CascadeRPNHead does not support test-time augmentation')

mmdet/models/dense_heads/centernet_head.py

@@ -10,10 +10,11 @@
 from ..utils.gaussian_target import (get_local_maximum, get_topk_from_heatmap,
  transpose_and_gather_feat)
 from .base_dense_head import BaseDenseHead
+from .dense_test_mixins import BBoxTestMixin
 
 
 @HEADS.register_module()
-class CenterNetHead(BaseDenseHead):
+class CenterNetHead(BaseDenseHead, BBoxTestMixin):
  """Objects as Points Head. CenterHead use center_point to indicate object's
  position. Paper link <https://arxiv.org/abs/1904.07850>
 

mmdet/models/dense_heads/corner_head.py

@@ -14,6 +14,7 @@
  get_topk_from_heatmap,
  transpose_and_gather_feat)
 from .base_dense_head import BaseDenseHead
+from .dense_test_mixins import BBoxTestMixin
 
 
 class BiCornerPool(BaseModule):
@@ -80,7 +81,7 @@ def forward(self, x):
 
 
 @HEADS.register_module()
-class CornerHead(BaseDenseHead):
+class CornerHead(BaseDenseHead, BBoxTestMixin):
  """Head of CornerNet: Detecting Objects as Paired Keypoints.
 
  Code is modified from the `official github repo

mmdet/models/dense_heads/dense_test_mixins.py

@@ -1,42 +1,42 @@
+import sys
 from inspect import signature
 
 import torch
 
-from mmdet.core import bbox2result, bbox_mapping_back, multiclass_nms
+from mmdet.core import bbox_mapping_back, merge_aug_proposals, multiclass_nms
 
+if sys.version_info >= (3, 7):
+ from mmdet.utils.contextmanagers import completed
 
-class BBoxTestMixin:
- """Mixin class for test time augmentation of bboxes."""
 
- def merge_aug_bboxes(self, aug_bboxes, aug_scores, img_metas):
- """Merge augmented detection bboxes and scores.
+class BBoxTestMixin(object):
+ """Mixin class for testing det bboxes via DenseHead."""
+
+ def simple_test_bboxes(self, feats, img_metas, rescale=False):
+ """Test det bboxes without test-time augmentation, can be applied in
+ DenseHead except for RPNHead and its variants(GuidedRPN etc.)
 
  Args:
- aug_bboxes (list[Tensor]): shape (n, 4*#class)
- aug_scores (list[Tensor] or None): shape (n, #class)
- img_shapes (list[Tensor]): shape (3, ).
+ feats (tuple[torch.Tensor]): Multi-level features from the
+ upstream network, each is a 4D-tensor.
+ img_metas (list[dict]): List of image information.
+ rescale (bool, optional): Whether to rescale the results.
+ Defaults to False.
 
  Returns:
- tuple: (bboxes, scores)
+ list[tuple[Tensor, Tensor]]: Each item in result_list is 2-tuple.
+ The first item is bboxes with shape (n, 5), where 5 represent
+ (tl_x, tl_y, br_x, br_y, score).
+ The shape of the second tensor in the tuple is labels
+ with shape (n,)
  """
- recovered_bboxes = []
- for bboxes, img_info in zip(aug_bboxes, img_metas):
- img_shape = img_info[0]['img_shape']
- scale_factor = img_info[0]['scale_factor']
- flip = img_info[0]['flip']
- flip_direction = img_info[0]['flip_direction']
- bboxes = bbox_mapping_back(bboxes, img_shape, scale_factor, flip,
- flip_direction)
- recovered_bboxes.append(bboxes)
- bboxes = torch.cat(recovered_bboxes, dim=0)
- if aug_scores is None:
- return bboxes
- else:
- scores = torch.cat(aug_scores, dim=0)
- return bboxes, scores
+ outs = self.forward(feats)
+ results_list = self.get_bboxes(*outs, img_metas, rescale=rescale)
+ return results_list
 
  def aug_test_bboxes(self, feats, img_metas, rescale=False):
- """Test det bboxes with test time augmentation.
+ """Test det bboxes with test time augmentation, can be applied in
+ DenseHead except for RPNHead and its variants(GuidedRPN etc.)
 
  Args:
  feats (list[Tensor]): the outer list indicates test-time
@@ -49,7 +49,11 @@ def aug_test_bboxes(self, feats, img_metas, rescale=False):
  Defaults to False.
 
  Returns:
- list[ndarray]: bbox results of each class
+ list[tuple[Tensor, Tensor]]: Each item in result_list is 2-tuple.
+ The first item is bboxes with shape (n, 5), where 5 represent
+ (tl_x, tl_y, br_x, br_y, score).
+ The shape of the second tensor in the tuple is labels
+ with shape (n,), Length of list should always be 1.
  """
  # check with_nms argument
  gb_sig = signature(self.get_bboxes)
@@ -96,5 +100,96 @@ def aug_test_bboxes(self, feats, img_metas, rescale=False):
  _det_bboxes = det_bboxes.clone()
  _det_bboxes[:, :4] *= det_bboxes.new_tensor(
  img_metas[0][0]['scale_factor'])
- bbox_results = bbox2result(_det_bboxes, det_labels, self.num_classes)
- return bbox_results
+
+ return [
+ (_det_bboxes, det_labels),
+ ]
+
+ def simple_test_rpn(self, x, img_metas):
+ """Test without augmentation, only for RPNHead and its
+ variants(GuidedRPN etc.)
+
+ Args:
+ x (tuple[Tensor]): Features from the upstream network, each is
+ a 4D-tensor.
+ img_metas (list[dict]): Meta info of each image.
+
+ Returns:
+ list[Tensor]: Proposals of each image, each item has shape (n, 5),
+ where 5 represent (tl_x, tl_y, br_x, br_y, score).
+ """
+ rpn_outs = self(x)
+ proposal_list = self.get_bboxes(*rpn_outs, img_metas)
+ return proposal_list
+
+ def aug_test_rpn(self, feats, img_metas):
+ """Test with augmentation for only for RPNHead and
+ its variants(GuidedRPN etc.)
+ Args:
+ feats (tuple[Tensor]): Features from the upstream network, each is
+ a 4D-tensor.
+ img_metas (list[dict]): Meta info of each image.
+
+ Returns:
+ list[Tensor]: Proposals of each image, each item has shape (n, 5),
+ where 5 represent (tl_x, tl_y, br_x, br_y, score).
+ """
+ samples_per_gpu = len(img_metas[0])
+ aug_proposals = [[] for _ in range(samples_per_gpu)]
+ for x, img_meta in zip(feats, img_metas):
+ proposal_list = self.simple_test_rpn(x, img_meta)
+ for i, proposals in enumerate(proposal_list):
+ aug_proposals[i].append(proposals)
+ # reorganize the order of 'img_metas' to match the dimensions
+ # of 'aug_proposals'
+ aug_img_metas = []
+ for i in range(samples_per_gpu):
+ aug_img_meta = []
+ for j in range(len(img_metas)):
+ aug_img_meta.append(img_metas[j][i])
+ aug_img_metas.append(aug_img_meta)
+ # after merging, proposals will be rescaled to the original image size
+ merged_proposals = [
+ merge_aug_proposals(proposals, aug_img_meta, self.test_cfg)
+ for proposals, aug_img_meta in zip(aug_proposals, aug_img_metas)
+ ]
+ return merged_proposals
+
+ if sys.version_info >= (3, 7):
+
+ async def async_simple_test_rpn(self, x, img_metas):
+ sleep_interval = self.test_cfg.pop('async_sleep_interval', 0.025)
+ async with completed(
+ __name__, 'rpn_head_forward',
+ sleep_interval=sleep_interval):
+ rpn_outs = self(x)
+
+ proposal_list = self.get_bboxes(*rpn_outs, img_metas)
+ return proposal_list
+
+ def merge_aug_bboxes(self, aug_bboxes, aug_scores, img_metas):
+ """Merge augmented detection bboxes and scores.
+
+ Args:
+ aug_bboxes (list[Tensor]): shape (n, 4*#class)
+ aug_scores (list[Tensor] or None): shape (n, #class)
+ img_shapes (list[Tensor]): shape (3, ).
+
+ Returns:
+ tuple: (bboxes, scores)
+ """
+ recovered_bboxes = []
+ for bboxes, img_info in zip(aug_bboxes, img_metas):
+ img_shape = img_info[0]['img_shape']
+ scale_factor = img_info[0]['scale_factor']
+ flip = img_info[0]['flip']
+ flip_direction = img_info[0]['flip_direction']
+ bboxes = bbox_mapping_back(bboxes, img_shape, scale_factor, flip,
+ flip_direction)
+ recovered_bboxes.append(bboxes)
+ bboxes = torch.cat(recovered_bboxes, dim=0)
+ if aug_scores is None:
+ return bboxes
+ else:
+ scores = torch.cat(aug_scores, dim=0)
+ return bboxes, scores

mmdet/models/dense_heads/detr_head.py

@@ -680,3 +680,29 @@ def _get_bboxes_single(self,
  det_bboxes = torch.cat((det_bboxes, scores.unsqueeze(1)), -1)
 
  return det_bboxes, det_labels
+
+ def simple_test_bboxes(self, feats, img_metas, rescale=False):
+ """Test det bboxes without test-time augmentation.
+
+ Args:
+ feats (tuple[torch.Tensor]): Multi-level features from the
+ upstream network, each is a 4D-tensor.
+ img_metas (list[dict]): List of image information.
+ rescale (bool, optional): Whether to rescale the results.
+ Defaults to False.
+
+ Returns:
+ list[tuple[Tensor, Tensor]]: Each item in result_list is 2-tuple.
+ The first item is bboxes with shape (n, 5), where 5 represent
+ (tl_x, tl_y, br_x, br_y, score).
+ The shape of the second tensor in the tuple is labels
+ with shape (n,)
+ """
+ batch_size = len(img_metas)
+ assert batch_size == 1, 'Currently only batch_size 1 for inference ' \
+ f'mode is supported. Found batch_size {batch_size}.'
+
+ # forward of this head requires img_metas
+ outs = self.forward(feats, img_metas)
+ results_list = self.get_bboxes(*outs, img_metas, rescale=rescale)
+ return results_list

mmdet/models/dense_heads/embedding_rpn_head.py

@@ -105,3 +105,11 @@ def forward_train(self, img, img_metas):
  def simple_test_rpn(self, img, img_metas):
  """Forward function in testing stage."""
  return self._decode_init_proposals(img, img_metas)
+
+ def simple_test(self, img, img_metas):
+ """Forward function in testing stage."""
+ raise NotImplementedError
+
+ def aug_test_rpn(self, feats, img_metas):
+ raise NotImplementedError(
+ 'EmbeddingRPNHead does not support test-time augmentation')

mmdet/models/dense_heads/ga_rpn_head.py

@@ -9,11 +9,10 @@
 
 from ..builder import HEADS
 from .guided_anchor_head import GuidedAnchorHead
-from .rpn_test_mixin import RPNTestMixin
 
 
 @HEADS.register_module()
-class GARPNHead(RPNTestMixin, GuidedAnchorHead):
+class GARPNHead(GuidedAnchorHead):
  """Guided-Anchor-based RPN head."""
 
  def __init__(self,

mmdet/models/dense_heads/paa_head.py

@@ -533,7 +533,9 @@ def _get_bboxes(self,
  cls_scores. Besides, score voting is used when `` score_voting``
  is set to True.
  """
- assert with_nms, 'PAA only supports "with_nms=True" now'
+ assert with_nms, 'PAA only supports "with_nms=True" now and it ' \
+ 'means PAAHead does not support ' \
+ 'test-time augmentation'
  assert len(cls_scores) == len(bbox_preds) == len(mlvl_anchors)
  batch_size = cls_scores[0].shape[0]
 

mmdet/models/dense_heads/rpn_head.py

@@ -7,11 +7,10 @@
 
 from ..builder import HEADS
 from .anchor_head import AnchorHead
-from .rpn_test_mixin import RPNTestMixin
 
 
 @HEADS.register_module()
-class RPNHead(RPNTestMixin, AnchorHead):
+class RPNHead(AnchorHead):
  """RPN head.
 
  Args: