Commits · 2b53953096b183f184b6dba2f210c1f92b1666ae · OpenDAS / OpenPCDet

29 Jul, 2020 1 commit
- continune to update cuda codes to support PyTorch 1.1/1.5 · 4db49780
  Shaoshuai Shi authored Jul 29, 2020
  
  4db49780
28 Jul, 2020 1 commit
- fix seed for sample_points data_processsor, update configs · d2d32f69
  Shaoshuai Shi authored Jul 28, 2020
  
  d2d32f69
27 Jul, 2020 2 commits
- bugfixed to test with multi-class nms on multihead · 3b2bcdaf
  Shaoshuai Shi authored Jul 25, 2020
  
  3b2bcdaf
- support multi-classes nms for multi-head, not checked · a236428b
  Shaoshuai Shi authored Jul 25, 2020
  
  a236428b
24 Jul, 2020 1 commit
- support PartA2_free models, update PointHead · 8075b170
  Shaoshuai Shi authored Jul 24, 2020
  
  8075b170
23 Jul, 2020 3 commits
- support PointRCNNHead with roipoint_pool3d, and pointrcnn.yaml · 29ad7c89
  Shaoshuai Shi authored Jul 23, 2020
  
  29ad7c89
- support PointNet2MSG backbone_3d for PointRCNN · a90d9375
  Shaoshuai Shi authored Jul 23, 2020
  
  a90d9375
- add RETURN_ENCODED_TENSOR config for Spconv UNet · e6f6151c
  Shaoshuai Shi authored Jul 23, 2020
  
  e6f6151c
22 Jul, 2020 7 commits
- bugfixed: set num_point_features without vfe, typos (#177) · b492d113
  Shaoshuai Shi authored Jul 23, 2020
  
  b492d113
- bugfixed: remove batch_index after proposal_layer · 0d524bc1
  Shaoshuai Shi authored Jul 23, 2020
  
  0d524bc1
- add config DISABLE_PART for PartA2Head · 202d767d
  Shaoshuai Shi authored Jul 23, 2020
  
  202d767d
- support PointRCNN with PartA2 head · 8ec2a2d4
  Shaoshuai Shi authored Jul 23, 2020
  
  8ec2a2d4
- support dense_head PointHeadBox · cf29d887
  Shaoshuai Shi authored Jul 23, 2020
  
  cf29d887
- support PointNet2Backbone · 0b21d8f9
  Shaoshuai Shi authored Jul 23, 2020
  
  0b21d8f9
- bugfixed: set num_point_features without vfe, typos · a2e7d474
  Shaoshuai Shi authored Jul 23, 2020
  
  a2e7d474
21 Jul, 2020 1 commit
- bugfixed: shape check in vsa layer and nms (#174) · 9babdfda
  Shaoshuai Shi authored Jul 21, 2020
  
  9babdfda
16 Jul, 2020 1 commit
- add torch.no_grad context for proposal_layer (#160) · 9f5d201e
  Shaoshuai Shi authored Jul 16, 2020
  
  9f5d201e
13 Jul, 2020 4 commits
- optimize the target_assigner to be more efficient · 7ac5625a
  Shaoshuai Shi authored Jul 13, 2020
  
  7ac5625a
- speed up PFNLayer · 69326cef
  Shaoshuai Shi authored Jul 13, 2020
  
  69326cef
- add config cbgs_1conv_sepreg_dir.yaml · f70902b8
  Shaoshuai Shi authored Jul 13, 2020
  
  f70902b8
- speed up the axis_aligned_target_assigner.py · eaa36cef
  Shaoshuai Shi authored Jul 13, 2020
  
  eaa36cef
12 Jul, 2020 2 commits
- fix tiny bugs · 83753361
  Shaoshuai Shi authored Jul 12, 2020
  
  83753361
- bugfixed: DO NOT encode angle by sin when using cossin angle encoding · f93948af
  Shaoshuai Shi authored Jul 12, 2020
  
  f93948af
11 Jul, 2020 2 commits
- bugfixed: separte regression heads training · ba28c08f
  Shaoshuai Shi authored Jul 11, 2020
  
  ba28c08f
- support separate regression heads for different properties · 72088ee3
  Shaoshuai Shi authored Jul 11, 2020
  
  72088ee3
09 Jul, 2020 1 commit
- set pos_cls_weight=1.0, neg_cls_weight=2.0 for NuScenes configs · cb32ff13
  Shaoshuai Shi authored Jul 09, 2020
  
  cb32ff13
08 Jul, 2020 1 commit
- add pp_multihead.yaml config for NuScenes, modify loc weights from 2.0 to 0.25 · 1bc21f70
  Shaoshuai Shi authored Jul 08, 2020
  
  1bc21f70
07 Jul, 2020 1 commit
- refactoring some configs and codes of multi-head · eb074cf2
  Shaoshuai Shi authored Jul 07, 2020
  
  eb074cf2
06 Jul, 2020 9 commits
- bugfixed: indice_key in VoxelResBackBone8x · 8f4ec1e3
  Shaoshuai Shi authored Jul 06, 2020
  
  8f4ec1e3
- bugfixed: loss calculation for separted multi-head · 0ac2901c
  Shaoshuai Shi authored Jul 06, 2020
  
  0ac2901c
- support more flexible setting of conv head; slice inputs when batch size is... · 19c66f79
  Gus-Guo authored Jul 04, 2020
```
support more flexible setting of conv head; slice inputs when batch size is too large in PFNLayer to avoid bugs (#124)

* support more flexible setting

* slice inputs of nn.Linear when batch size is too large
```
  19c66f79
- bugfixed: typo in anchor_head_multi · 6b3e52c9
  Shaoshuai Shi authored Jul 06, 2020
  
  6b3e52c9
- support multiheads that predict seperate classes · b16e0891
  Gus-Guo authored Jul 06, 2020
  
  b16e0891
- add VoxelResBackbone8x 3D backbone · 7d251220
  Shaoshuai Shi authored Jul 06, 2020
  
  7d251220
- support WeightedL1Loss in anchor_head_template · d0489cf6
  Shaoshuai Shi authored Jul 06, 2020
  
  d0489cf6
- bugfixed: run nuscene testing with velocity · ea6cf247
  Shaoshuai Shi authored Jul 04, 2020
  
  ea6cf247
- support velocity prediction with SECOND in NuScenes · c1d93158
  Shaoshuai Shi authored Jul 03, 2020
  
  c1d93158
04 Jul, 2020 1 commit

support more flexible setting of conv head; slice inputs when batch size is... · b42d7a99

Gus-Guo authored Jul 04, 2020

support more flexible setting of conv head; slice inputs when batch size is too large in PFNLayer to avoid bugs (#124)

* support more flexible setting

* slice inputs of nn.Linear when batch size is too large

b42d7a99

03 Jul, 2020 2 commits
- simplify codes of axis aligned target assigner and update corresponding config · 4337fc4e
  Gus-Guo authored Jul 03, 2020
  
  4337fc4e
- simplify codes of multihead and update multihead config · 3cd5c47b
  Gus-Guo authored Jul 03, 2020
  
  3cd5c47b