Contour Knowledge Transfer for Salient Object Detection

by Xin Li, Fan Yang, Hong Cheng, Wei Liu and Dinggang Shen

Introduction

This repository is for 'Contour Knowledge Transfer for Salient ObjectDetection'.

Installation

For installation, please follow the instructions of Caffe.

The code has been tested successfully on Ubuntu 14.04 with CUDA 8.0.

Usage

1. Clone the repository:

   git clone https://github.com/lixin666/C2SNet.git

2. Build Caffe and pycaffe:

   cd caffe-master
   cp Makefile.config.example Makefile.config
   vim Makefile.config
   make -j8 && make pycaffe

ps: You should uncomment 'WITH_PYTHON_LAYER := 1' in Makefile.config before compiling.

3. Test:

   • Test code is in folder 'code'.

   • We provide two models trained with 10K (MSRA10K) and 30K (MSRA10K + Web Images) training images. Download trained models and put them in folder 'code/models':

   • C2SNet10K.caffemodel: BaiduYun or GoogleDrive

   • C2SNet30K.caffemodel: BaiduYun or GoogleDrive

   • Put the test images in folder 'images', and run

   python run_demo.py

   -After that, the results will be genrated in folder 'res'.

4. Results:

• You can also download the results for comparison.

• C2SNet10K: BaiduYun(access code: p3b7) or GoogleDrive

• C2SNet30K: BaiduYun(access code: 2c67) or GoogleDrive

Citation

If C2SNet is useful for your research, please consider citing:

@inproceedings{xin2018c2s,
  author = {Li, Xin and Yang, Fan and  Cheng, Hong, and Liu, Wei and Shen, Dinggang},
  title = {Contour Knowledge Transfer for Salient Object Detection},
  booktitle = {ECCV},
  year = {2018}
}

Question

Please contact 'xinli_uestc@hotmail.com' Or 'fanyang_uestc@hotmail.com'