Eli-YiLi / PMM

Pseudo-mask Matters in Weakly-supervised Semantic Segmentation

Pseudo-mask Matters in Weakly-supervised Semantic Segmentation

By Yi Li, Zhanghui Kuang, Liyang Liu, Yimin Chen, Wayne Zhang

SenseTime, Tsinghua University

Table of Contents

0. Introduction
1. Classification
2. Segmentation
3. License

Introduction

This is a PyTorch implementation of Pseudo-mask Matters in Weakly-supervised Semantic Segmentation.(ICCV2021).

In this paper, we propose Coefficient of Variation Smoothing and Proportional Pseudo-mask Generation to generate high quality pseudo-mask in classification part. In segmentation part, we propose Pretended Under-Fitting strategy and Cyclic Pseudo-mask for better utilization of pseudo-mask.

Classification

Data Preparation

1. Download VOC12 OneDrive, BaiduYun
2. Download COCO14 BaiduYun
3. Download pretrained models OneDrive, BaiduYun

(extract code of BaiduYun: mtci)

Get Started

git clone https://github.com/Eli-YiLi/PMM
cd PMM
ln -s [path to model files] models
ln -s [path to VOC12] voc12
ln -s [path to COCO14] coco14
pip3 install -r requirements.txt
bash slurm_run.sh [partition name] [dataset name] / bash dist_run.sh [dataset name]

Segmentation

Please refer to WSSS_MMSeg

License

Please refer to: LICENSE.