SVAM: Saliency-guided Visual Attention Modeling (To Appear at RSS 2022)

Pointers

• Paper: http://www.roboticsproceedings.org/rss18/p048.pdf
• Preprint: https://arxiv.org/pdf/2011.06252.pdf
• Video demonstration: https://youtu.be/SxJcsoQw7KI
• Data: http://irvlab.cs.umn.edu/resources/usod-dataset
• Project page: http://irvlab.cs.umn.edu/visual-attention-modeling/svam

SVAM-Net Model

• Jointly accommodate bottom-up and top-down learning in two branches sharing the same encoding layers
• Incorporates four spatial attention modules (SAMs) along these learning pathways
• Exploits coarse-level and fine-level semantic features for SOD at four stages of abstractions
• The bottom-up pipeline (SVAM-Net_Light) performs abstract saliency prediction at fast rates
• The top-down pipeline ensures fine-grained saliency estimation by aresidual refinement module (RRM)
• Pretrained weights can be downloaded from this Google-Drive link

SVAM-Net Features

• Provides SOTA performance for SOD on underwater imagery
• Exhibits significantly better generalization performance than existing solutions
• Achieves fast end-to-end inference
  • The end-to-end SVAM-Net : 20.07 FPS in GTX-1080, 4.5 FPS on Jetson Xavier
  • Decoupled SVAM-Net_Light: 86.15 FPS in GTX-1080, 21.77 FPS on Jetson Xavier

USOD Dataset

• A new challenging test set for benchmark evaluation of underwater SOD models
• Contains 300 natural underwater images and ground truth labels
• Can be downloaded from: http://irvlab.cs.umn.edu/resources/usod-dataset
• Evaluation code: https://github.com/xahidbuffon/SOD-Evaluation-Tool-Python
• Evaluation data can be found in this Google-Drive link

Bibliography entry:

@inproceedings{islam2022svam,
author={Islam, Md Jahidul and Wang, Ruobing and Sattar, Junaed},
title={{SVAM: Saliency-guided Visual Attention Modeling 
    	    by Autonomous Underwater Robots}},
booktitle={Robotics: Science and Systems (RSS)},
year={2022},
address={NY, USA}
}

Acknowledgements

• https://github.com/CaitinZhao/cvpr2019_Pyramid-Feature-Attention-Network-for-Saliency-detection
• https://github.com/Ugness/PiCANet-Implementation
• https://github.com/wenguanwang/SODsurvey
• https://github.com/wenguanwang/PAGE-Net
• https://github.com/backseason/PoolNet
• https://github.com/wenguanwang/ASNet
• https://github.com/NathanUA/BASNet
• https://github.com/wuzhe71/CPD