fix max_voxels

open-mmlab · Apr 26, 2023 · 8e6c33e · 8e6c33e
1 parent b900f2d
commit 8e6c33e
Show file tree

Hide file tree

Showing 3 changed files with 9 additions and 8 deletions.
diff --git a/configs/_base_/models/minkunet.py b/configs/_base_/models/minkunet.py
@@ -9,15 +9,14 @@
  point_cloud_range=[-100, -100, -20, 100, 100, 20],
  voxel_size=[0.05, 0.05, 0.05],
  max_voxels=(-1, -1)),
- ),
+  max_voxels=80000),
  backbone=dict(
  type='MinkUNetBackbone',
  in_channels=4,
  base_channels=32,
- encoder_channels=[32, 64, 128, 256],
- decoder_channels=[256, 128, 96, 96],
  num_stages=4,
- init_cfg=None),
+ encoder_channels=[32, 64, 128, 256],
+ decoder_channels=[256, 128, 96, 96]),
  decode_head=dict(
  type='MinkUNetHead',
  channels=96,

diff --git a/configs/_base_/models/spvcnn.py b/configs/_base_/models/spvcnn.py
@@ -9,14 +9,14 @@
  point_cloud_range=[-100, -100, -20, 100, 100, 20],
  voxel_size=[0.05, 0.05, 0.05],
  max_voxels=(-1, -1)),
- ),
+  max_voxels=80000),
  backbone=dict(
  type='SPVCNNBackbone',
  in_channels=4,
  base_channels=32,
+ num_stages=4,
  encoder_channels=[32, 64, 128, 256],
  decoder_channels=[256, 128, 96, 96],
- num_stages=4,
  drop_ratio=0.3),
  decode_head=dict(
  type='MinkUNetHead',

diff --git a/mmdet3d/models/data_preprocessors/data_preprocessor.py b/mmdet3d/models/data_preprocessors/data_preprocessor.py
@@ -49,6 +49,8 @@ class Det3DDataPreprocessor(DetDataPreprocessor):
  voxelization and dynamic voxelization. Defaults to 'hard'.
  voxel_layer (dict or :obj:`ConfigDict`, optional): Voxelization layer
  config. Defaults to None.
+ max_voxels (int): Maximum number of voxels in each voxel grid. Defaults
+ to None.
  mean (Sequence[Number], optional): The pixel mean of R, G, B channels.
  Defaults to None.
  std (Sequence[Number], optional): The pixel standard deviation of
@@ -77,7 +79,7 @@ def __init__(self,
  voxel: bool = False,
  voxel_type: str = 'hard',
  voxel_layer: OptConfigType = None,
- max_voxels: int = 80000,
+ max_voxels: Optional[int] = None,
  mean: Sequence[Number] = None,
  std: Sequence[Number] = None,
  pad_size_divisor: int = 1,
@@ -428,7 +430,7 @@ def voxelize(self, points: List[torch.Tensor],
  inds, point2voxel_map = self.sparse_quantize(
  res_coors_numpy, return_index=True, return_inverse=True)
  point2voxel_map = torch.from_numpy(point2voxel_map).cuda()
- if self.training:
+ if self.training and self.max_voxels is not None:
  if len(inds) > self.max_voxels:
  inds = np.random.choice(
  inds, self.max_voxels, replace=False)