/csg-km

IV 2020 "CSG: Critical Scenario Generation from Real Traffic Accidents"

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

Critical Scenario Generation Toolkit

Introduction

This repo offers the source code for Critical Scnenario Generation (CSG) toolkit. The toolkit aims to extract dynamic (trajectories) and static (road) elements from a given surveillance or dash camera video, and fromulated as OpenDrive and OpenScenario files as outputs.

How to build

  1. Install system dependencies:
sudo apt-get install python3-tk
  1. Install python3 package:
pip3 install Cython numpy
pip3 install -r requirements.txt
  1. compile cython package
cd src/track/kcf && make & cd ../../../

How to run:

  1. Edit variables in config.py (Optional)
  2. Download machine learning models from Google Drive and unzip to "data" folder.
  3. Run
  • For survelliance video:
python3 CSG.py
  • For dash camera video:

It uses ORB-SLAM2 to get extrinsic parameters of moving camera. Please install ORB-SLAM2 first and put the executable file under the "/src/SLAM" folder.

Then run:

python3 CSG_ego.py

Reqruiments

At least 1 GPU is needed. By default, the models are deployed on gpu:0. You can change your settings in "config" file.

Folder structure

+-- critical-scenario-generation
¦   +-- data
¦   ¦   +-- yolo
¦   ¦   +-- vehicle_reid
¦   ¦   +-- mask_rcnn_ego
¦   ¦   +-- mask_rcnn
¦   ¦   +-- lane
¦   ¦   +-- depth
¦   +-- src
¦   ¦   +-- SLAM
¦   ¦   ¦   +-- mono_kitti
¦   ¦   +-- ...
¦   +-- Lib_OpenSCENARIO
¦   +-- icon
¦   +-- CSG.py
¦   +-- CSG_ego.py
¦   +-- ScaleSolver.py
¦   +-- config.py
¦   +-- requirements.txt

Licence

check LICENSE

Citation

If you use our source code, please consider citing the following:

@InProceedings{csg2020,
  title={CSG: critical scenario generation from real traffic accidents},
  author={Zhang, Xinxin and Li, Fei and Wu, Xiangbin},
  booktitle = {IV},
  year={2020}
}