AmressamAI / AlphaRefine

Official implementation for Alpha-Refine

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Alpha-Refine: Boosting Tracking Performance by Precise Bounding Box Estimation

This repo provides the official implementation of AlphaRefine tracker, which wins the VOT2020 Real-Time Challenge. The implementation of our latest arxiv paper will also be released here.

News

  • 🏆 Alpha-Refine wins VOT2020 Real-Time Challenge with EAOMultistart 0.499!
  • 2020.08.29 Our presentation slide has been uploaded.

Run Alpha-Refine on VOT2020

The code for VOT2020 participation has been uploaded to AlphaRefine_submit directory. Please following README.md under that directory to run our tracker.

Introduction

Alpha-Refine is an accurate, fast, and general refinement module. When combined with existing base trackers, Alpha-Refine can significantly boost their tracking performance, in no need of joint training or fine-tuning.

Architecture

The overall architecture is shown below

Video Demos

More video demos can be found at Google Drive, Baidu(extraction code:q5fs)

About

Official implementation for Alpha-Refine


Languages

Language:Python 99.2%Language:Jupyter Notebook 0.7%Language:MATLAB 0.1%Language:Shell 0.0%