Unofficial implementation of LEDNet(https://github.com/xiaoyufenfei/LEDNet) on ADE20K

Introduction

This project contains the code of LEDNet supporting ADE20K dataset ( LEDNet: A Lightweight Encoder-Decoder Network for Real-time Semantic Segmentation by Yu Wang)

LEDNet is a lightweight network, with an asymmetric encoder-decoder architecture for the task of real-time semantic segmentation.

Training

• Download the ADE20K scene parsing dataset:

cd datasets
chmod +x download_ADE20K.sh
./download_ADE20K.sh

• Adjusting parameter before training, such as GPU ID, learning rate……

cd train
python train.py

Testing

• the trained models will be coming soon

cd train
python  test.py

Results

• coming soon

Reference

1. LEDNet
2. awesome-semantic-segmentation-pytorch