Skip to content

Latest commit

 

History

History
28 lines (17 loc) · 716 Bytes

README.md

File metadata and controls

28 lines (17 loc) · 716 Bytes

Introduction

This repo is an unofficial implementation of IoU Loss for 2D/3D Object Detection. It contains the simple calculattion of IoUs of 2D / 3D rotated bounding box.

Requirements

Following dependencies are needed

cudatoolkit=10.2
pytorch>1.5         
numpy
matplotlib

Usage

First, compile the CUDA extension.

cd cuda_op
python setup.py install

Then, run a demo which validate the Pytorch functions and CUDA extension.

cd ..
python IoUDemo.py

Future Work

Complete the implementation of GIoU-loss and DIoU-loss for rotated bounding boxes.