PPEDNet: Pyramid Pooling Encoder-Decoder Network for Real-Time Semantic Segmentation

Introduction

PPEDNet is an encoder-decoder netwrok based on VGG16 framework for image semantic segmentation. This work focuses on the tradeoff between performance and speed.

Usage

Supported OS: the source code was tested on 64-bit Ubuntu 14.04.3 LTS Linux OS, and it should also be executable in other linux distributions.

Installation: Please compile the modified Caffe framework caffe-segnet (It supports all all necessaty layers for PPEDNet).

Pre-trained Model: Please download the VGG16 pre-traiend model (https://github.com/BVLC/caffe/wiki/Model-Zoo).

Citing Our Work

If you find PPEDNet useful in your research, please consider to cite our paper:

@inproceedings{ICIG2017PPEDNet,
   title={PPEDNet: Pyramid Pooling Encoder-Decoder Network for Real-Time Semantic Segmentation},
   author={Zhentao Tan, Bin Liu, and Nenghai Yu},
   booktitle={ICIG},
   year={2017}
}

All rights reserved.