Commits · f561b8ae288f3b6fa6e5aca971ebbcd3012c18b3 · OpenDAS / OpenPCDet · GitLab

28 Jul, 2020 1 commit
- bugfixed: backward bug in ThreeInterpolate of stacked version pointnet2 · f561b8ae
  Shaoshuai Shi authored Jul 28, 2020
  
  f561b8ae
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 5 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 PointNet2 batch version from https://github.com/sshaoshuai/Pointnet2.PyTorch · adbb322f
  Shaoshuai Shi authored Jul 23, 2020
  
  adbb322f
- add sample_points data_processor from PointRCNN · 30df59ed
  Shaoshuai Shi authored Jul 23, 2020
  
  30df59ed
- add RETURN_ENCODED_TENSOR config for Spconv UNet · e6f6151c
  Shaoshuai Shi authored Jul 23, 2020
  
  e6f6151c
22 Jul, 2020 9 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
- support PointNet2 3D backbone with pointnet2_stack cuda codes · 6b4dae4d
  Shaoshuai Shi authored Jul 23, 2020
  
  6b4dae4d
- support PointResidualCoder · 8447a475
  Shaoshuai Shi authored Jul 23, 2020
  
  8447a475
- bugfixed: set num_point_features without vfe, typos · a2e7d474
  Shaoshuai Shi authored Jul 23, 2020
  
  a2e7d474
21 Jul, 2020 2 commits
- bugfixed: use relative path in nuscene infos for sweeps · 1c4e1391
  Shaoshuai Shi authored Jul 21, 2020
  
  1c4e1391
- 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
15 Jul, 2020 1 commit
- Make browsing TensorBoard easier (#152) · 2400fdf2
  Martin Hahner authored Jul 15, 2020
```
* Make TensorBoard more concise

* Export only metrics using 40 recalll positions
```
  2400fdf2
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 3 commits
- fix tiny bugs · efb94221
  Shaoshuai Shi authored Jul 12, 2020
  
  efb94221
- 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 4 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
- bugfixed: continue to fix velocity flip and rot · 3a551ac1
  Shaoshuai Shi authored Jul 11, 2020
  
  3a551ac1
- bugfixed: flip and rot velocity when doing augmentation · 14ac4d12
  Shaoshuai Shi authored Jul 11, 2020
  
  14ac4d12
09 Jul, 2020 6 commits
- set pos_cls_weight=1.0, neg_cls_weight=2.0 for NuScenes configs · cb32ff13
  Shaoshuai Shi authored Jul 09, 2020
  
  cb32ff13
- bugfixed: support merge_all_iters_to_one_epoch in dataloader · 0dd5326e
  Shaoshuai Shi authored Jul 09, 2020
  
  0dd5326e
- bugfixed: support merge_all_iters_to_one_epoch in dataloader (#142) · b7553c87
  Shaoshuai Shi authored Jul 09, 2020
```
* bugfixed: support merge_all_iters_to_one_epoch in dataloader
```
  b7553c87
- bugfixed: support merge_all_iters_to_one_epoch in dataloader · f50c306a
  Shaoshuai Shi authored Jul 09, 2020
  
  f50c306a
- support merge_all_iters_to_one_epoch in NuScenes · 7ef14309
  Shaoshuai Shi authored Jul 09, 2020
  
  7ef14309
- add config SET_NAN_VELOCITY_TO_ZEROS=True for NuScenes · 9157ee44
  Shaoshuai Shi authored Jul 09, 2020
  
  9157ee44
08 Jul, 2020 1 commit
- support merge_all_iters_to_one_epoch to avoid deadlock when switching epochs,... · 67b43a3b
  Shaoshuai Shi authored Jul 08, 2020
```
support merge_all_iters_to_one_epoch to avoid deadlock when switching epochs, eval last 10 epochs after training
```
  67b43a3b