Electrical Engineering and Systems Science > Image and Video Processing
[Submitted on 15 Apr 2021 (v1), revised 26 Jun 2021 (this version, v2), latest version 10 Sep 2021 (v3)]
Title:BAM: A Lightweight and Efficient Balanced Attention Mechanism for Single Image Super Resolution
View PDFAbstract:Attention mechanism has shown enormous potential for single image super-resolution (SISR). However, existing works only proposed some attention mechanism for a specific network. A universal attention mechanism for SISR, which could further improve the performance of networks without attention and provide a baseline for networks with attention, is still lacking. To fit this gap, we propose a lightweight and efficient Balanced Attention Mechanism (BAM), which consists of Avgpool Channel Attention Module (ACAM) and Maxpool Spatial Attention Module (MSAM) in parallel. The information extraction mechanism of ACAM and MSAM effectively filters redundant information, making the overall structure of BAM very lightweight. Owing to the parallel structure, during the gradient backpropagation process of BAM, ACAM and MSAM not only conduct self-optimization, but also mutual optimization so as to generate more balanced attention information. To verify the effectiveness and robustness of BAM, we applied it to 12 state-ofthe-art SISR networks. The results on 4 benchmark datasets demonstrate that BAM can efficiently improve the networks' performance, and for those with attention, the substitution with BAM further reduces the amount of parameters and increase the inference speed. Moreover, ablation experiments were conducted to prove the minimalism of BAM.
Submission history
From: Fanyi Wang [view email][v1] Thu, 15 Apr 2021 16:22:16 UTC (1,314 KB)
[v2] Sat, 26 Jun 2021 08:00:21 UTC (1,995 KB)
[v3] Fri, 10 Sep 2021 13:09:47 UTC (15,370 KB)
Current browse context:
eess.IV
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.