The official codes for "Adaptive Fusion Affinity Graph with Noise-free Online Low-rank Representation for Natural Image Segmentation" The AFA-Graph paper can be found here. A list of papers and datasets about natural/color image segmentation can be found in our repository.
AFA-Graph is modified from the offcial SAS implementation.
The code requires the version of Matlab2018a, Ubuntu 16.04.
The original MSRC dataset can be downloaded from here. We place the processed dataset in database/MSRC/ folder.
Run the demo IKDE_BSDS300_searchN.m for 'BSD300'.
Run the demo IKDE_BSDS500_searchN.m for 'BSD500'.
Run the demo IKDE_MSRC_searchN.m for 'MSRC'.
Run the demo IKDE_SBD_searchN.m for 'SBD'.
Run the demo IKDE_VOC_searchN.m for 'PASCAL VOC'.
The detailed results of BSD300 can be found in results_searchN_denoise_n/ikde_1_1/evaluation_BSD300.txt.
The detailed results of BSD500 can be found in results_searchN_denoise_n/ikde_1_1/evaluation_BSD500.txt.
The detailed results of MSRC can be found in results_searchN_denoise_n/ikde_1_1/evaluation_MSRC.txt.
The detailed results of SBD can be found in results_searchN_denoise_n/ikde_1_1/evaluation_SBD.txt.
The detailed results of VOC can be found in results_searchN_denoise_n/ikde_1_1/evaluation_VOC.txt.
If you find this repository useful in your research, please consider citing:
@INPROCEEDINGS{AFA-Graph,
author={Y. {Zhang} and M. {Liu} and H. {Zhang} and G. {Sun} and J. {He}},
booktitle={arXiv:2110.11685},
title={Adaptive Fusion Affinity Graph with Noise-free Online Low-rank Representation for Natural Image Segmentation},
year={2021}}