Monocular 3D Object Detection: An Extrinsic Parameter Free Approach (CVPR2021)

Yunsong Zhou, Yuan He, Hongzi Zhu, Cheng Wang, Hongyang Li, Qinhong Jiang

Our paper is now avaiable on CVPR 2021 open access.

Introduction

Our framework is implemented and tested with Ubuntu 16.04, CUDA 8.0/9.0, Python 3, Pytorch 0.4/1.0/1.1, NVIDIA Tesla V100/TITANX GPU.

If you find our work useful in your research please consider citing our paper:

@InProceedings{Zhou_2021_CVPR,
author    = {Zhou, Yunsong and He, Yuan and Zhu, Hongzi and Wang, Cheng and Li, Hongyang and Jiang, Qinhong},
title     = {Monocular 3D Object Detection: An Extrinsic Parameter Free Approach},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month     = {June},
year      = {2021},
pages     = {7556-7566}
}

Requirements

• Cuda & Cudnn & Python & Pytorch

  This project is tested with CUDA 8.0/9.0, Python 3, Pytorch 0.4/1.0/1.1, NVIDIA Tesla V100/TITANX GPU. And almost all the packages we use are covered by Anaconda.

  Please install proper CUDA and CUDNN version, and then install Anaconda3 and Pytorch.

Data preparation

Download and unzip the full KITTI detection dataset.

Training

I am currently busy with my own courses. I will sort out the work involved in the near future. Relevant code and models will be avaiable soon.