det3

This repo is a LiDAR-based 3D object detection code base. It provides some basic operations and methods for 3D object detection. The architecture of the det3 project is as follows:

Requirements

• python 3.7.3
• CUDA 10.1+
• PyTorch 1.4.0+
• Open3D 0.9

Dockerfile

cd dockerfiles
docker build . -t <dockerimage-tag>/det3:v0.1
docker run -it --name det3 --gpus all <dockerimage-tag>/det3:v0.1

Usage

You can call the functions of det3 as a package. It provides basic functions in det3.dataloader, det3.ops, and det3.visualizer with detailed documentation inline. The inline documentation of each function explains itself.

The det3 also provides a use case of SECONDNet in det3/methods/second. You can refer to this implementation for its usage.

We will also provide more detailed tutorials in the future.

Acknowledgement

• second.pytorch
• spconv
• kitt_eval

Thanks for the contribution of predecessors in this field. If you find any inappropriate references in this code base, please feel free to contact us (pyun@ust.hk).