Home

Awesome

PWC PWC PWC PWC

<div align="center">

<img src="./assets/lion.jpg" alt="Image 2" width="4%" style="margin: 0 auto;" > LION: Linear Group RNN for 3D Object Detection in Point Clouds

Zhe Liu <sup>1,* </sup>, Jinghua Hou <sup>1,* </sup>, Xinyu Wang <sup>1,* </sup>, Xiaoqing Ye <sup>3</sup>, Jingdong Wang <sup>3</sup>, Hengshuang Zhao <sup>2</sup>, Xiang Bai <sup>1,✉</sup> <br> <sup>1</sup> Huazhong University of Science and Technology, <sup>2</sup> The University of Hong Kong, <sup>3</sup> Baidu Inc. <br> * Equal contribution, ✉ Corresponding author. <br>

Project Page | NeurIPS 2024

<img src="./assets/all.jpg" alt="Image 2" width="60%" style="margin: 0 auto;" > </div>

🔥 Highlights

News

Results

ModelmAP/mAPH_L1mAP/mAPH_L2Vec_L1Vec_L2Ped_L1Ped_L2Cyc_L1Cyc_L2Config
LION-RetNet80.9/78.874.6/72.779.0/78.570.6/70.284.6/80.077.2/72.879.0/78.076.1/75.1config
LION-RWKV81.0/79.074.7/72.879.7/79.371.3/71.084.6/80.077.1/72.778.7/77.775.8/74.8config
LION-Mamba81.4/79.475.1/73.279.5/79.171.1/70.784.9/80.477.5/73.279.7/78.776.7/75.8config
LION-Mamba-L82.1/80.175.9/74.080.3/79.972.0/71.685.8/81.478.5/74.380.1/79.077.2/76.2config

Note: You could reduce the training epochs from 24 to 12~(the performance gap is within 1 mAP/mAPH) or reduce the 100% training to 20% training sets.

ModelSplitEpochCBGSNDSmAPConfigDownload (Baidu Pan)Download (Google Drive)
LION-RetNetVal36False71.967.3confignus_retnet.pth (ksmp)nus_retnet.pth
LION-RWKVVal36False71.766.8config
LION-MambaVal36False72.168.0confignus_mamba.pth (2tvc)nus_mamba.pth
LION-MambaVal48False72.368.2config
LION-MambaTest36False73.969.8

Note: Our model on nuScenes does not use CBGS for training more time and without any test-time augmentation or model ensembling! For obtaining more stable and better performance, you could try to train more time~(e.g., 48 epochs)

ModelmAPConfigDownload (Baidu Pan)Download (Google Drive)
LION-RetNet40.7configargov2_retnet.pth (yghm)argov2_retnet.pth
LION-RWKV41.1configargov2_rwkv.pth (cr4e)argov2_rwkv.pth
LION-Mamba41.5configargov2_mamba.pth (k63i)argov2_mamba.pth
ModelVehiclePedestrianCyclistmAPConfigDownload
LION-RetNet78.152.468.366.3config
LION-RWKV78.350.668.465.8config
LION-Mamba78.253.268.566.6config

Quick Validation

ModelCarPedestrianCyclistConfigDownload
LION-TTT78.058.669.6config
LION-xLSTM77.959.367.4config
LION-RetNet77.960.269.6config
LION-Mamba78.360.268.6config
LION-RWKV78.362.271.2config

Installation

Please refer to INSTALL.md for the installation of LION codebase.

Getting Started

We provide all training&evaluation scripts for training our LION, please refer to tools/

bash run_train_lion_for_nus.sh
bash run_train_lion_for_waymo.sh
bash run_train_lion_for_argov2.sh
bash run_train_lion_for_once.sh
bash run_train_lion_for_kitti.sh

For more details about LION, please refer to GETTING_STARTED.md to learn more usage about LION.

TODO

Citation

@article{liu2024lion,
  title={LION: Linear Group RNN for 3D Object Detection in Point Clouds},
  author={Zhe Liu, Jinghua Hou, Xingyu Wang, Xiaoqing Ye, Jingdong Wang, Hengshuang Zhao, Xiang Bai},
  journal={Advances in Neural Information Processing Systems},
  year={2024}
  }

Acknowledgements

We thank these great works and open-source repositories: OpenPCDet, DSVT, FlatFormer, HEDNet, Mamba, RWKV, Vision-RWKV, RMT, xLSTM, TTT, and flash-linear-attention.