-
New dataset API (unstable during alpha), almost completely remove kitti-specific code. you can add your custom dataset by following steps: (1): implement all Dataset API functions (2): use web visualization tool to check whether the box is correct. (3): add your dataset to all_dataset.py, change the dataset_class_name in config file.
-
Add NuScenes dataset support (incomplete in 1.6.0alpha), I plan to reproduce the NDS score in their paper.
-
Add pointpillars to this repo.
-
Full Tensorboard support.
-
FP16 and multi-gpu (need test, I only have one gpu) support.
-
Move all data-specific functions to their corresponding dataset file.
-
Improved config file structure, remove some unused item.
-
remove much unused and deprecated code.
-
add two learning rate scheduler: exp decay and manual step
- Better support for custom lidar data. You need to check KittiDataset for more details. (no test yet, I don't have custom data)
- Change all box to center format.
- Change kitti info format, now you need to regenerate kitti infos and gt database.
- Eval functions now support custom data evaluation. you need to specify z_center and z_axis in eval function.
- Better RPN, you can add custom block by inherit RPNBase and implement _make_layer method.
- Update pretrained model.
- Add a simple inference notebook. everyone should start this project by that notebook.
- Add windows support. Training on windows is slow than linux.
- New sparse convolution based models. VFE-based old models are deprecated. Now the model looks like this: points([N, 4])->voxels([N, 5, 4])->Features([N, 4])->Sparse Convolution Networks->RPN. See this for more details of sparse conv networks.
- The SparseConvNet is deprecated. New library spconv is introduced.
- Super converge (from fastai) is implemented. Now all network can converge to a good result with only 50~80 epoch. For example.
car.fhd.configonly needs 50 epochs to reach 78.3 AP (car mod 3d). - Target assigner now works correctly when using multi-class.