Source code for the paper "An Adaptive Affinity Graph with Subspace Pursuit for Natural Image Segmentation". The AASP-Graph paper can be found here.
AASP-Graph is modified from the offcial GL-Graph implementation.
The code requires the version of Matlab2018a, Ubuntu 16.04.
The BSDS300 dataset and predata (extracted features) can be downloaded from GL-Graph and place them in BSD/ and bsd_300_feat/ folder, respectively.
Run the demo demo_AASP_BSDS300.m.
The detailed results can be found in BSDS300_evaluation.txt. More relevant results of natural image segmentation can be found in our projects.
If you find this repository useful in your research, please consider citing:
@INPROCEEDINGS{AASP-Graph,
author={Y. {Zhang} and H. {Zhang} and Y. {Guo} and K. {Lin} and J. {He}},
booktitle={IEEE International Conference on Multimedia and Expo (ICME)},
title={An Adaptive Affinity Graph with Subspace Pursuit for Natural Image Segmentation},
year={2019}，
pages={802-807},}