Awesome
OPGP
This repo is the implementation of:
Occupancy Prediction-Guided Neural Planner for Autonomous Driving <br> Haochen Liu, Zhiyu Huang, Chen Lv <br> AutoMan Research Lab, Nanyang Technological University <br> [Paper] [arXiv] [Zhihu]
- Code is now released 😀!
Overview
In this repository, you can expect to find the following features 🤩:
- Pipelines for data process and training
- Open-loop evaluations
Not included 😵:
- Model weights (Due to license from WOMD)
- Real-time planning (Codes are not optimized for real-time performance)
Experiment Pipelines
Dataset and Environment
-
Downloading Waymo Open Motion Dataset v1.1. Utilize data from
scenario/training_20s
for train set, and data fromscenario/validation
for val & test. -
Clone this repository and install required packages.
-
[NOTED] For theseus library, you may build from scratch and add system PATH in
planner.py
Data Process
- Preprocess data for training & testing:
python preprocess.py \
--root_dir path/to/your/Waymo_Dataset/scenario/ \
--save_dir path/to/your/processed_data/ \
--processes=16
- You may also refer to Waymo_candid_list for more interactive and safety-critical scenarios filtered in
scenario/validation
Training & Testing
- Train & Eval the model using the command:
python -m torch.distributed.launch \
--nproc_per_node 1 \ # number of gpus
--master_port 16666 \
training.py \
--data_dir path/to/your/processed_data/ \
--save_dir path/to/save/your/logs/
- Conduct Open-loop Testing using the command:
python testing.py \
--data_dir path/to/your/testing_data/ \
--model_dir path/to/pretrained/model/
Citation
If you find this repository useful for your research, please consider giving us a star 🌟 and citing our paper.
@inproceedings{liu2023occupancy,
title={Occupancy prediction-guided neural planner for autonomous driving},
author={Liu, Haochen and Huang, Zhiyu and Lv, Chen},
booktitle={2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC)},
pages={4859--4865},
year={2023},
organization={IEEE}
}