disprcnn

Code release for Stereo 3D Object Detection via Shape Prior Guided Instance Disparity Estimation (CVPR 2020).

This project contains the implementation of our CVPR 2020 paper arxiv.

Authors: Jiaming Sun, Linghao Chen, Yiming Xie, Siyu Zhang, Qinhong Jiang, Xiaowei Zhou, Hujun Bao.

Requirements

  • Ubuntu 16.04+
  • Python 3.7+
  • 8 Nvidia GPU with mem >= 12G (recommended, see Notes for details.)
  • GCC >= 4.9
  • PyTorch 1.2.0

Install

# Install webp support
sudo apt install libwebp-dev
# Clone repo
git clone https://github.com/zju3dv/disprcnn.git
cd disprcnn
# Install conda environment
conda env create -f environment.yaml
conda activate disprcnn
# Install Disp R-CNN
sh build_and_install.sh

Training and evaluation

See TRAIN_VAL.md

Sample results

qualitative

Citation

If you find this code useful for your research, please use the following BibTeX entry.

@inproceedings{sun2020disprcnn,
  title={Disp R-CNN: Stereo 3D Object Detection via Shape Prior Guided Instance Disparity Estimation},
  author={Sun, Jiaming and Chen, Linghao and Xie, Yiming and Zhang, Siyu and Jiang, Qinhong and Zhou, Xiaowei and Bao, Hujun},
  booktitle={CVPR},
  year={2020}
}

Acknowledgment

This repo is built based on the Mask R-CNN implementation from maskrcnn-benchmark, and we also use the pretrained Stereo R-CNN weight from here for initialization.

License

Copyright (c) 2020 3D Vision Group of State Key Lab at CAD&CG, Zhejiang University

GitHub