CFAT Triangular-Attention Super-Resolution — Poster
A conference poster presenting CFAT, a novel image super-resolution method using triangular window attention. It details the architecture, ablation studies, and quantitative/qualitative results against state-of-the-art methods like SwinIR and HAT.
Paper context
Paper title: CFAT: Unleashing Triangular Windows for Image Super-resolution Abstract: A conference poster presenting CFAT, a novel image super-resolution method using triangular window attention. It details the architecture, ablation studies, and quantitative/qualitative results against state-of-the-art methods like SwinIR and HAT. Paper body (method & results): <!DOCTYPE html> <html lang="en"> <head> <meta content="text/html; charset=utf-8" http-equiv="content-type"/> <title>CFAT: Unleashing Triangular Windows for Image Super-resolution</title> <!--Generated on Sun Mar 24 12:56:46 2024 by LaTeXML (version 0.8.7) http://dlmf.nist.gov/LaTeXML/.--> <meta content="width=device-width, initial-scale=1, shrink-to-fit=no" name="viewport"/> <link href="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/css/bootstrap.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/ar5iv_0.7.4.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/latexml_styles.css" rel="stylesheet" type="text/css"/> <script src="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/js/bootstrap.bundle.min.js"></script> <script src="https://cdnjs.cloudflare.com/ajax/libs/html2canvas/1.3.3/html2canvas.min.js"></script> <script src="/static/browse/0.3.4/js/addons.js"></script> <script src="/static/browse/0.3.4/js/feedbackOverlay.js"></script> <base href="/html/2403.16143v1/"/></head> <body> <nav class="ltx_page_navbar"> <nav class="ltx_TOC"> <ol class="ltx_toclist"> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S1" title="1 Introduction ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">1 </span>Introduction</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S2" title="2 Related Works ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">2 </span>Related Works</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S2.SS1" title="2.1 CNN Based Super-Resolution ‣ 2 Related Works ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">2.1 </span>CNN Based Super-Resolution</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S2.SS2" title="2.2 Vision Transformer (ViT) Based SR ‣ 2 Related Works ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">2.2 </span>Vision Transformer (ViT) Based SR</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S3" title="3 Proposed Method ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3 </span>Proposed Method</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S3.SS1" title="3.1 Overall Architecture ‣ 3 Proposed Method ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.1 </span>Overall Architecture</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S3.SS2" title="3.2 Head Module - Shallow Feature Extractor: ‣ 3 Proposed Method ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.2 </span>Head Module - Shallow Feature Extractor:</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"> <a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S3.SS3" title="3.3 Body Module - Deep Feature Extractor: ‣ 3 Proposed Method ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.3 </span>Body Module - Deep Feature Extractor:</span></a> <ol class="ltx_toclist ltx_toclist_subsection"> <li class="ltx_tocentry ltx_tocentry_subsubsection"> <a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S3.SS3.SSS1" title="3.3.1 Dense Window Attention Blocks (DWAB) ‣ 3.3 Body Module - Deep Feature Extractor: ‣ 3 Proposed Method ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.3.1 </span>Dense Window Attention Blocks (DWAB)</span></a> <ol class="ltx_toclist ltx_toclist_subsubsection"> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S3.SS3.SSS1.Px1" title="Dense-Hybrid Window Attention Block (D-HWAB): ‣ 3.3.1 Dense Window Attention Blocks (DWAB) ‣ 3.3 Body Module - Deep Feature Extractor: ‣ 3 Proposed Method ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title">Dense-Hybrid Window Attention Block (D-HWAB):</span></a></li> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S3.SS3.SSS1.Px2" title="Overlaping Cross Fusion Attention Block (OCFAB): ‣ 3.3.1 Dense Window Attention Blocks (DWAB) ‣ 3.3 Body Module - Deep Feature Extractor: ‣ 3 Proposed Method ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title">Overlaping Cross Fusion Attention Block (OCFAB):</span></a></li> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S3.SS3.SSS1.Px3" title="Channe-Wise Attention Block (CWAB): ‣ 3.3.1 Dense Window Attention Blocks (DWAB) ‣ 3.3 Body Module - Deep Feature Extractor: ‣ 3 Proposed Method ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title">Channe-Wise Attention Block (CWAB):</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_subsubsection"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S3.SS3.SSS2" title="3.3.2 Sparse Window Attention Block (SWAB) ‣ 3.3 Body Module - Deep Feature Extractor: ‣ 3 Proposed Method ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.3.2 </span>Sparse Window Attention Block (SWAB)</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsubsection"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S3.SS3.SSS3" title="3.3.3 Tail Module - HQ Image Reconstruction ‣ 3.3 Body Module - Deep Feature Extractor: ‣ 3 Proposed Method ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.3.3 </span>Tail Module - HQ Image Reconstruction</span></a></li> </ol> </li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S4" title="4 Triangular Window Technique ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4 </span>Triangular Window Technique</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S4.SS0.SSS0.Px1" title="Design: ‣ 4 Triangular Window Technique ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title">Design:</span></a></li> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S4.SS0.SSS0.Px2" title="Advantages over Rectangular Window: ‣ 4 Triangular Window Technique ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title">Advantages over Rectangular Window:</span></a></li> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S4.SS0.SSS0.Px3" title="Computational Cost: ‣ 4 Triangular Window Technique ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title">Computational Cost:</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S5" title="5 Experiments ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5 </span>Experiments</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S5.SS1" title="5.1 Datasets and Performance Matrices ‣ 5 Experiments ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.1 </span>Datasets and Performance Matrices</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S5.SS2" title="5.2 Experimental Settings ‣ 5 Experiments ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.2 </span>Experimental Settings</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"> <a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S5.SS3" title="5.3 Ablation Study ‣ 5 Experiments ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.3 </span>Ablation Study</span></a> <ol class="ltx_toclist ltx_toclist_subsection"> <li class="ltx_tocentry ltx_tocentry_subsubsection"> <a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S5.SS3.SSS1" title="5.3.1 Impact of Various Model Hyperparameters : ‣ 5.3 Ablation Study ‣ 5 Experiments ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.3.1 </span>Impact of Various Model Hyperparameters :</span></a> <ol class="ltx_toclist ltx_toclist_subsubsection"> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S5.SS3.SSS1.Px1" title="Window Size and Shift Size : ‣ 5.3.1 Impact of Various Model Hyperparameters : ‣ 5.3 Ablation Study ‣ 5 Experiments ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title">Window Size and Shift Size :</span></a></li> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2403.16143v1#S5.SS3.SSS1.Px2" title="Interval Size : ‣ 5.3.1 Impact of Various Model Hyperparameters : ‣ 5.3 Ablation Study ‣ 5 Experiments ‣ CFAT: Unleashing Triangular Windows for Image Super-resolution"><span class="ltx_text ltx_ref_title">Interval Size :</span></a></li> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https:/
