Update type hints on part 10.

mmdet/models/task_modules/coders/bucketing_bbox_coder.py

@@ -1,10 +1,14 @@
 # Copyright (c) OpenMMLab. All rights reserved.
+from typing import Optional, Sequence, Tuple, Union
+
 import numpy as np
 import torch
 import torch.nn.functional as F
+from torch import Tensor
 
 from mmdet.registry import TASK_UTILS
-from mmdet.structures.bbox import HorizontalBoxes, bbox_rescale, get_box_tensor
+from mmdet.structures.bbox import (BaseBoxes, HorizontalBoxes, bbox_rescale,
+ get_box_tensor)
 from .base_bbox_coder import BaseBBoxCoder
 
 
@@ -32,13 +36,13 @@ class BucketingBBoxCoder(BaseBBoxCoder):
  """
 
  def __init__(self,
- num_buckets,
- scale_factor,
- offset_topk=2,
- offset_upperbound=1.0,
- cls_ignore_neighbor=True,
- clip_border=True,
- **kwargs):
+ num_buckets: int,
+ scale_factor: int,
+ offset_topk: int = 2,
+ offset_upperbound: float = 1.0,
+ cls_ignore_neighbor: bool = True,
+ clip_border: bool = True,
+ **kwargs) -> None:
  super().__init__(**kwargs)
  self.num_buckets = num_buckets
  self.scale_factor = scale_factor
@@ -47,7 +51,8 @@ def __init__(self,
  self.cls_ignore_neighbor = cls_ignore_neighbor
  self.clip_border = clip_border
 
- def encode(self, bboxes, gt_bboxes):
+ def encode(self, bboxes: Union[Tensor, BaseBoxes],
+ gt_bboxes: Union[Tensor, BaseBoxes]) -> Tuple[Tensor]:
  """Get bucketing estimation and fine regression targets during
  training.
 
@@ -71,7 +76,12 @@ def encode(self, bboxes, gt_bboxes):
  self.cls_ignore_neighbor)
  return encoded_bboxes
 
- def decode(self, bboxes, pred_bboxes, max_shape=None):
+ def decode(
+ self,
+ bboxes: Union[Tensor, BaseBoxes],
+ pred_bboxes: Tensor,
+ max_shape: Optional[Tuple[int]] = None
+ ) -> Tuple[Union[Tensor, BaseBoxes], Tensor]:
  """Apply transformation `pred_bboxes` to `boxes`.
  Args:
  boxes (torch.Tensor or :obj:`BaseBoxes`): Basic boxes.
@@ -97,7 +107,9 @@ def decode(self, bboxes, pred_bboxes, max_shape=None):
  return bboxes, loc_confidence
 
 
-def generat_buckets(proposals, num_buckets, scale_factor=1.0):
+def generat_buckets(proposals: Tensor,
+ num_buckets: int,
+ scale_factor: float = 1.0) -> Tuple[Tensor]:
  """Generate buckets w.r.t bucket number and scale factor of proposals.
 
  Args:
@@ -145,13 +157,13 @@ def generat_buckets(proposals, num_buckets, scale_factor=1.0):
  return bucket_w, bucket_h, l_buckets, r_buckets, t_buckets, d_buckets
 
 
-def bbox2bucket(proposals,
- gt,
- num_buckets,
- scale_factor,
- offset_topk=2,
- offset_upperbound=1.0,
- cls_ignore_neighbor=True):
+def bbox2bucket(proposals: Tensor,
+ gt: Tensor,
+ num_buckets: int,
+ scale_factor: float,
+ offset_topk: int = 2,
+ offset_upperbound: float = 1.0,
+ cls_ignore_neighbor: bool = True) -> Tuple[Tensor]:
  """Generate buckets estimation and fine regression targets.
 
  Args:
@@ -268,13 +280,14 @@ def bbox2bucket(proposals,
  return offsets, offsets_weights, bucket_labels, bucket_cls_weights
 
 
-def bucket2bbox(proposals,
- cls_preds,
- offset_preds,
- num_buckets,
- scale_factor=1.0,
- max_shape=None,
- clip_border=True):
+def bucket2bbox(proposals: Tensor,
+ cls_preds: Tensor,
+ offset_preds: Tensor,
+ num_buckets: int,
+ scale_factor: float = 1.0,
+ max_shape: Optional[Union[Sequence[int], Tensor,
+ Sequence[Sequence[int]]]] = None,
+ clip_border: bool = True) -> Tuple[Tensor]:
  """Apply bucketing estimation (cls preds) and fine regression (offset
  preds) to generate det bboxes.
 

mmdet/models/task_modules/coders/delta_xywh_bbox_coder.py

@@ -1,11 +1,13 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 import warnings
+from typing import Optional, Sequence, Union
 
 import numpy as np
 import torch
+from torch import Tensor
 
 from mmdet.registry import TASK_UTILS
-from mmdet.structures.bbox import HorizontalBoxes, get_box_tensor
+from mmdet.structures.bbox import BaseBoxes, HorizontalBoxes, get_box_tensor
 from .base_bbox_coder import BaseBBoxCoder
 
 
@@ -32,20 +34,21 @@ class DeltaXYWHBBoxCoder(BaseBBoxCoder):
  """
 
  def __init__(self,
- target_means=(0., 0., 0., 0.),
- target_stds=(1., 1., 1., 1.),
- clip_border=True,
- add_ctr_clamp=False,
- ctr_clamp=32,
- **kwargs):
+ target_means: Sequence[float] = (0., 0., 0., 0.),
+ target_stds: Sequence[float] = (1., 1., 1., 1.),
+ clip_border: bool = True,
+ add_ctr_clamp: bool = False,
+ ctr_clamp: int = 32,
+ **kwargs) -> None:
  super().__init__(**kwargs)
  self.means = target_means
  self.stds = target_stds
  self.clip_border = clip_border
  self.add_ctr_clamp = add_ctr_clamp
  self.ctr_clamp = ctr_clamp
 
- def encode(self, bboxes, gt_bboxes):
+ def encode(self, bboxes: Union[Tensor, BaseBoxes],
+ gt_bboxes: Union[Tensor, BaseBoxes]) -> Tensor:
  """Get box regression transformation deltas that can be used to
  transform the ``bboxes`` into the ``gt_bboxes``.
 
@@ -65,11 +68,14 @@ def encode(self, bboxes, gt_bboxes):
  encoded_bboxes = bbox2delta(bboxes, gt_bboxes, self.means, self.stds)
  return encoded_bboxes
 
- def decode(self,
- bboxes,
- pred_bboxes,
- max_shape=None,
- wh_ratio_clip=16 / 1000):
+ def decode(
+ self,
+ bboxes: Union[Tensor, BaseBoxes],
+ pred_bboxes: Tensor,
+ max_shape: Optional[Union[Sequence[int], Tensor,
+ Sequence[Sequence[int]]]] = None,
+ wh_ratio_clip: Optional[float] = 16 / 1000
+ ) -> Union[Tensor, BaseBoxes]:
  """Apply transformation `pred_bboxes` to `boxes`.
 
  Args:
@@ -123,7 +129,12 @@ def decode(self,
  return decoded_bboxes
 
 
-def bbox2delta(proposals, gt, means=(0., 0., 0., 0.), stds=(1., 1., 1., 1.)):
+def bbox2delta(
+ proposals: Tensor,
+ gt: Tensor,
+ means: Sequence[float] = (0., 0., 0., 0.),
+ stds: Sequence[float] = (1., 1., 1., 1.)
+) -> Tensor:
  """Compute deltas of proposals w.r.t. gt.
 
  We usually compute the deltas of x, y, w, h of proposals w.r.t ground
@@ -168,15 +179,16 @@ def bbox2delta(proposals, gt, means=(0., 0., 0., 0.), stds=(1., 1., 1., 1.)):
  return deltas
 
 
-def delta2bbox(rois,
- deltas,
- means=(0., 0., 0., 0.),
- stds=(1., 1., 1., 1.),
- max_shape=None,
- wh_ratio_clip=16 / 1000,
- clip_border=True,
- add_ctr_clamp=False,
- ctr_clamp=32):
+def delta2bbox(rois: Tensor,
+ deltas: Tensor,
+ means: Sequence[float] = (0., 0., 0., 0.),
+ stds: Sequence[float] = (1., 1., 1., 1.),
+ max_shape: Optional[Union[Sequence[int], Tensor,
+ Sequence[Sequence[int]]]] = None,
+ wh_ratio_clip: float = 16 / 1000,
+ clip_border: bool = True,
+ add_ctr_clamp: bool = False,
+ ctr_clamp: int = 32) -> Tensor:
  """Apply deltas to shift/scale base boxes.
 
  Typically the rois are anchor or proposed bounding boxes and the deltas are
@@ -267,15 +279,16 @@ def delta2bbox(rois,
  return bboxes
 
 
-def onnx_delta2bbox(rois,
- deltas,
- means=(0., 0., 0., 0.),
- stds=(1., 1., 1., 1.),
- max_shape=None,
- wh_ratio_clip=16 / 1000,
- clip_border=True,
- add_ctr_clamp=False,
- ctr_clamp=32):
+def onnx_delta2bbox(rois: Tensor,
+ deltas: Tensor,
+ means: Sequence[float] = (0., 0., 0., 0.),
+ stds: Sequence[float] = (1., 1., 1., 1.),
+ max_shape: Optional[Union[Sequence[int], Tensor,
+ Sequence[Sequence[int]]]] = None,
+ wh_ratio_clip: float = 16 / 1000,
+ clip_border: Optional[bool] = True,
+ add_ctr_clamp: bool = False,
+ ctr_clamp: int = 32) -> Tensor:
  """Apply deltas to shift/scale base boxes.
 
  Typically the rois are anchor or proposed bounding boxes and the deltas are

mmdet/models/task_modules/coders/distance_point_bbox_coder.py

@@ -1,6 +1,10 @@
 # Copyright (c) OpenMMLab. All rights reserved.
+from typing import Optional, Sequence, Union
+
+from torch import Tensor
+
 from mmdet.registry import TASK_UTILS
-from mmdet.structures.bbox import (HorizontalBoxes, bbox2distance,
+from mmdet.structures.bbox import (BaseBoxes, HorizontalBoxes, bbox2distance,
  distance2bbox, get_box_tensor)
 from .base_bbox_coder import BaseBBoxCoder
 
@@ -17,11 +21,15 @@ class DistancePointBBoxCoder(BaseBBoxCoder):
  border of the image. Defaults to True.
  """
 
- def __init__(self, clip_border=True, **kwargs):
+ def __init__(self, clip_border: Optional[bool] = True, **kwargs) -> None:
  super().__init__(**kwargs)
  self.clip_border = clip_border
 
- def encode(self, points, gt_bboxes, max_dis=None, eps=0.1):
+ def encode(self,
+ points: Tensor,
+ gt_bboxes: Union[Tensor, BaseBoxes],
+ max_dis: Optional[float] = None,
+ eps: float = 0.1) -> Tensor:
  """Encode bounding box to distances.
 
  Args:
@@ -41,7 +49,13 @@ def encode(self, points, gt_bboxes, max_dis=None, eps=0.1):
  assert gt_bboxes.size(-1) == 4
  return bbox2distance(points, gt_bboxes, max_dis, eps)
 
- def decode(self, points, pred_bboxes, max_shape=None):
+ def decode(
+ self,
+ points: Tensor,
+ pred_bboxes: Tensor,
+ max_shape: Optional[Union[Sequence[int], Tensor,
+ Sequence[Sequence[int]]]] = None
+ ) -> Union[Tensor, BaseBoxes]:
  """Decode distance prediction to bounding box.
 
  Args:

mmdet/models/task_modules/coders/legacy_delta_xywh_bbox_coder.py

@@ -1,9 +1,12 @@
 # Copyright (c) OpenMMLab. All rights reserved.
+from typing import Optional, Sequence, Union
+
 import numpy as np
 import torch
+from torch import Tensor
 
 from mmdet.registry import TASK_UTILS
-from mmdet.structures.bbox import HorizontalBoxes, get_box_tensor
+from mmdet.structures.bbox import BaseBoxes, HorizontalBoxes, get_box_tensor
 from .base_bbox_coder import BaseBBoxCoder
 
 
@@ -32,14 +35,15 @@ class LegacyDeltaXYWHBBoxCoder(BaseBBoxCoder):
  """
 
  def __init__(self,
- target_means=(0., 0., 0., 0.),
- target_stds=(1., 1., 1., 1.),
- **kwargs):
+ target_means: Sequence[float] = (0., 0., 0., 0.),
+ target_stds: Sequence[float] = (1., 1., 1., 1.),
+ **kwargs) -> None:
  super().__init__(**kwargs)
  self.means = target_means
  self.stds = target_stds
 
- def encode(self, bboxes, gt_bboxes):
+ def encode(self, bboxes: Union[Tensor, BaseBoxes],
+ gt_bboxes: Union[Tensor, BaseBoxes]) -> Tensor:
  """Get box regression transformation deltas that can be used to
  transform the ``bboxes`` into the ``gt_bboxes``.
 
@@ -60,11 +64,14 @@ def encode(self, bboxes, gt_bboxes):
  self.stds)
  return encoded_bboxes
 
- def decode(self,
- bboxes,
- pred_bboxes,
- max_shape=None,
- wh_ratio_clip=16 / 1000):
+ def decode(
+ self,
+ bboxes: Union[Tensor, BaseBoxes],
+ pred_bboxes: Tensor,
+ max_shape: Optional[Union[Sequence[int], Tensor,
+ Sequence[Sequence[int]]]] = None,
+ wh_ratio_clip: Optional[float] = 16 / 1000
+ ) -> Union[Tensor, BaseBoxes]:
  """Apply transformation `pred_bboxes` to `boxes`.
 
  Args:
@@ -91,10 +98,12 @@ def decode(self,
  return decoded_bboxes
 
 
-def legacy_bbox2delta(proposals,
- gt,
- means=(0., 0., 0., 0.),
- stds=(1., 1., 1., 1.)):
+def legacy_bbox2delta(
+ proposals: Tensor,
+ gt: Tensor,
+ means: Sequence[float] = (0., 0., 0., 0.),
+ stds: Sequence[float] = (1., 1., 1., 1.)
+) -> Tensor:
  """Compute deltas of proposals w.r.t. gt in the MMDet V1.x manner.
 
  We usually compute the deltas of x, y, w, h of proposals w.r.t ground
@@ -139,12 +148,14 @@ def legacy_bbox2delta(proposals,
  return deltas
 
 
-def legacy_delta2bbox(rois,
- deltas,
- means=(0., 0., 0., 0.),
- stds=(1., 1., 1., 1.),
- max_shape=None,
- wh_ratio_clip=16 / 1000):
+def legacy_delta2bbox(rois: Tensor,
+ deltas: Tensor,
+ means: Sequence[float] = (0., 0., 0., 0.),
+ stds: Sequence[float] = (1., 1., 1., 1.),
+ max_shape: Optional[
+ Union[Sequence[int], Tensor,
+ Sequence[Sequence[int]]]] = None,
+ wh_ratio_clip: float = 16 / 1000) -> Tensor:
  """Apply deltas to shift/scale base boxes in the MMDet V1.x manner.
 
  Typically the rois are anchor or proposed bounding boxes and the deltas are