Skip to content

Latest commit

 

History

History
36 lines (22 loc) · 680 Bytes

readme.md

File metadata and controls

36 lines (22 loc) · 680 Bytes

RRNet: A Hybrid Detector for Object Detection in Drone-captured Images

Changrui Chen, Yu Zhang, Qingxuan LV, Shuo Wei, andd Xiaorui Wang

This repo is created for ICCV Drones Detection 2019. We proposed RRNet, which is an anchor-free detector with a re-regression module.

🎉 We are the Honorable Mention to the challenge!

Architecture

Demo

Environment

  • PyTorch 1.1.0
  • python3+

Setup

cd ext/nms && make && cd ..

# Prepare Drones dataset in 'data' folder.

# For Training RRNet
cp scripts/RRNet/train.py ./
python3 train.py

# For evaluation
cp scripts/RRNet/eval.py ./
python3 eval.py