Skip to content
Code for ICCV2019 paper "InstaBoost: Boosting Instance Segmentation Via Probability Map Guided Copy-Pasting"
Python Cuda C C++ JavaScript Shell Other
Branch: master
Clone or download
Latest commit 7ef9f98 Sep 10, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
detectron Update Sep 10, 2019
mmdetection add new models Aug 23, 2019
yolact update yolact, fix minor errors Aug 30, 2019 add new models Aug 23, 2019 Update Aug 31, 2019


This repository is implementation of ICCV2019 paper "InstaBoost: Boosting Instance Segmentation Via Probability Map Guided Copy-Pasting". Our paper has been released on arXiv

Install InstaBoost

To install InstaBoost, use this command.

pip install instaboost

The detail implementation can be found here.

Because InstaBoost depends on matting package here, we highly recommend users to use python3.5 or 3.6, OpenCV 2.4 to avoid some errors. Envrionment setting instructions can be found here.

Quick Start

Video demo for InstaBoost:

Currently we have integrated InstaBoost into three open implementations: mmdetection, detectron and yolact.

Since these frameworks may continue updating, codes in this repo may be a little different from their current repo.

Use InstaBoost In Your Project

It is easy to integrate InstaBoost into your framework. You can refer to instructions of our implementations on mmdetection, detectron and yolact

Setup InstaBoost Configurations

To change InstaBoost Configurations, users can use function InstaBoostConfig.

Model Zoo

Results and models are available in the Model zoo. More models are coming!


If you use this toolbox or benchmark in your research, please cite this project.

author = {Fang, Hao-Shu and Sun, Jianhua and Wang, Runzhong and Gou, Minghao and Li, Yong-Lu and Lu, Cewu},
title = {InstaBoost: Boosting Instance Segmentation Via Probability Map Guided Copy-Pasting},
journal={arXiv preprint arXiv:1908.07801},
year = {2019}

Please also cite mmdetection, detectron and yolact if you use the corresponding codes.


Our detection and instance segmentation framework is based on mmdetecion, detectron and yolact.

You can’t perform that action at this time.