Complex-YOLO

Complex-YOLO: Real-time 3D Object Detection on Point Clouds pytorch

Introduction

This is an unofficial implementation of Complex-YOLO: Real-time 3D Object Detection on Point Clouds in pytorch. A large part of this project is based on the work here:https://github.com/marvis/pytorch-yolo2

Point Cloud Preprocessing is based on:https://github.com/skyhehe123/VoxelNet-pytorch https://github.com/dongwoohhh/MV3D-Pytorch

Data Preparation

Download the 3D KITTI detection dataset.

Camera calibration matrices of object data set (16 MB)

Training labels of object data set (5 MB)

Velodyne point clouds (29 GB)

Train

python3 main.py

trained model(using DarkNet) download link :https://pan.baidu.com/s/1yeU3Q-Oyozv7qFqrVhzj0A

Result