Salient Object Detection via Integrity Learning
Authors: Mingchen Zhuge, Deng-Ping Fan, Nian Liu, Dingwen Zhang, Dong Xu, & Ling Shao.
This repository contains the source code, prediction results, and evaluation toolbox of our ICON method. The technical report could be found at [Paper][中文版].
Introduction
To facilitate integrity learning for SOD, we design a novel Integrity Cognition Network (ICON), which explores three important components to learn strong integrity features. 1) Unlike the existing models that focus more on feature discriminability, we introduce a diverse feature aggregation (DFA) component to aggregate features with various receptive fields (i.e., kernel shape and context) and increase the feature diversity. Such diversity is the foundation for mining the integral salient objects. 2) Based on the DFA features, we introduce the integrity channel enhancement (ICE) component with the goal of enhancing feature channels that highlight the integral salient objects (i.e., micro and macro levels) while suppressing the other distracting ones. 3) After extracting the enhanced features, the part-whole verification (PWV) method is employed to determine whether the part and whole object features have strong agreement. Such part-whole agreements can further improve the micro-level integrity for each salient object.
News
-
it supports VGG, ResNet, PVT, Swin, CycleMLP as alternative encoders.
-
it ensembles the quick SOD evaltools.
-
it releases an efficient way to conduct the SOC dataset.
-
it gives a code for drawing PR/F-measre curves.
Get Start
1. Download Datasets and Checkpoints.
- Datasets:
Baidu | 提取码:ICON or Goole Drive
- Checkpoints:
Baidu | 提取码:ICON or Goole Drive
2. Train ICON.
sh util/scripts/train_icon.sh
3. Test ICON.
sh util/scripts/test_icon.sh
4. Eval ICON.
sh util/scripts/run_sod_eval.sh
sh util/scripts/run_soc_eval.sh
Prediction Maps
- ICON-S saliency maps: Baidu | 提取码:ICON
- ICON-P saliency maps: Baidu | 提取码:ICON
- ICON-R saliency maps: Baidu | 提取码:ICON
- ICON-V saliency maps: Baidu | 提取码:ICON
Acknowlegement
Thanks Dawn-bin pointing out a code mistake. And thanks Jing-Hui Shi contributes a MLP-encoder version.
Citation
@article{zhuge2021salient,
title={Salient Object Detection via Integrity Learning},
author={Zhuge, Mingchen and Fan, Deng-Ping and Liu, Nian and Zhang, Dingwen and Xu, Dong and Shao, Ling},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
year={2022},
publisher={IEEE}
}