[ICRA 2024] RiskBench: A Scenario-based Benchmark for Risk Identification

---

Setup

git clone --depth 1 https://github.com/HCIS-Lab/RiskBench.git

System Requirements

• Linux ( Tested on Ubuntu 18.04, 20.04 )
• Python3 ( Tested on Python 3.7 )
• PyTorch ( Tested on PyTorch 1.10.0 )
• CUDA ( Tested on CUDA 11.3 )
• CARLA ( Tested on CARLA 0.9.14 )
• GPU ( Tested on Nvidia RTX3090, RTX4090 )
• CPU ( Tested on AMD 7950X3D, Intel 12900kf )

Dataset

• The complete RiskBench dataset is available for download

  • split folder into multiple large zip files.
  • split folder into multiple fixed tar files, each 2 Gb. .

• We provide DATA_FOR_Planning_Aware_Metric and DATASET_for_LBC_Training for planning aware metric evaluation and LBC training data respectively.

• We provide instructions on how to collect basic scenarios and data augmentation. Please refer to link.

---

Risk Identification Benchmark

Baseline

We provide each baseline's training and inference details which can be found here.

Offline Risk Identification Evaluation

We perform offline risk identification evaluation and fine-grained scenario-based analysis by taking input as preserved risk identification prediction. You can generate by following the instruction in this page.

Planning-aware Evaluation

We provide data collection pipeline and planning aware evluation platform which can be found here

---

Citation

@inproceedings{kung2024riskbench,
  title={RiskBench: A Scenario-based Benchmark for Risk Identification},
  author={Kung, Chi-Hsi and Yang, Chieh-Chi and Pao, Pang-Yuan and Lu, Shu-Wei and Chen, Pin-Lun and Lu, Hsin-Cheng and Chen, Yi-Ting},
  booktitle={2024 IEEE International Conference on Robotics and Automation (ICRA)},
  year={2024},
  organization={IEEE}
}