Awesome

Safety-aware Motion Prediction with Unseen Vehicles for Autonomous Driving

<a href="https://arxiv.org/abs/2109.01510"><img src="https://img.shields.io/badge/arXiv-2109.01510-b31b1b.svg"></a> <a href="https://opensource.org/licenses/MIT"><img src="https://img.shields.io/badge/License-MIT-yellow.svg"></a>

Safety-aware Motion Prediction with Unseen Vehicles for Autonomous Driving <br> Xuanchi Ren, Tao Yang, Li Erran Li, Alexandre Alahi, and Qifeng Chen<br> ICCV 2021<br>

[Paper] [Supplementary material]

Recent Updates

:white_check_mark: Update data preprocessing code
:white_check_mark: Update model
:black_square_button: Training script

Installation

Cloning

1. Clone this repository with the following command:

git clone https://github.com/xrenaa/Safety-Aware-Motion-Prediction.git
cd experiments/nuScenes
git clone https://github.com/nutonomy/nuscenes-devkit.git
git checkout 12fb09169eb8ebf04bc39a30cd50334215769c3e

2. Replace experiments/nuScenes/nuscenes-devkit/python-sdk/nuscenes/prediction/input_representation/static_layers.py with the file Here.

Environment Setup

First, we'll create a conda environment to hold the dependencies.

conda create --name safeDrive python=3.6 -y
source activate safeDrive
pip install -r requirements.txt

Data Setup

nuScenes Dataset

1. Download the nuScenes dataset (this requires signing up on their website). Extract the downloaded zip file's contents and place them in the experiments/nuScenes directory. Then, download the map expansion pack (v1.2) and copy the contents of the extracted maps folder into the experiments/nuScenes/maps folder. Eventually you should have the following folder structure:

experiments/nuscenes
    samples	        -	Sensor data for keyframes.
    sweeps	        -	Sensor data for intermediate frames.
    maps	        -	Folder for all map files: rasterized .png images and vectorized .json files.
    v1.0-trainval	-	JSON tables that include all the meta data and annotations.
    process_data.py	-	Our provided data processing script.

2. Finally, process them into a data format that our model can work with.

cd experiments/nuScenes

# For the tranval nuScenes dataset, use the following
python process_data.py --data ../nuScenes --split train --img_size 128
python process_data.py --data ../nuScenes --split train_val --img_size 128
python process_data.py --data ../nuScenes --split val --img_size 128

We provide a notebook to visualize the processed data.

Citation

@inproceedings{ren2021unseen,
  title   = {Safety-aware Motion Prediction with Unseen Vehicles for Autonomous Driving},
  author  = {Ren, Xuanchi, and Yang, Tao, and Li, Li Erran, and Alahi, Alexandre, and Chen, Qifeng},
  booktitle = {ICCV},
  year    = {2021}
}