MASA: Label-Free Any-Object Tracking — Poster
A conference poster presenting MASA, a method for generalizing Multiple Object Tracking (MOT) to any objects without tracking labels. It features model architecture diagrams, training and inference pipelines, quantitative results on benchmarks like TAO and BDD, and qualitative comparisons including t-SNE visualizations.
논문 컨텍스트
Paper title: Matching Anything by Segmenting Anything (MASA) - CVPR 2024 Poster Abstract: A conference poster presenting MASA, a method for generalizing Multiple Object Tracking (MOT) to any objects without tracking labels. It features model architecture diagrams, training and inference pipelines, quantitative results on benchmarks like TAO and BDD, and qualitative comparisons including t-SNE visualizations. Paper body (method & results): <!DOCTYPE html> <html lang="en"> <head> <meta content="text/html; charset=utf-8" http-equiv="content-type"/> <title>Matching Anything by Segmenting Anything</title> <!--Generated on Thu Jun 6 16:16:54 2024 by LaTeXML (version 0.8.8) 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.0.7.9.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/ar5iv-fonts.0.7.9.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_new.js"></script> <script src="/static/browse/0.3.4/js/feedbackOverlay.js"></script> <base href="/html/2406.04221v1/"/></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/2406.04221v1#S1" title="In Matching Anything by Segmenting Anything"><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/2406.04221v1#S2" title="In Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">2 </span>Related Work</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/2406.04221v1#S2.SS1" title="In 2 Related Work ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">2.1 </span>Learning Instance-level Association</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S2.SS2" title="In 2 Related Work ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">2.2 </span>Segment and Track Anything Models</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S3" title="In Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3 </span>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/2406.04221v1#S3.SS1" title="In 3 Method ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.1 </span>Preliminaries: SAM</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"> <a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S3.SS2" title="In 3 Method ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.2 </span>Matching Anything by Segmenting Anything</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/2406.04221v1#S3.SS2.SSS1" title="In 3.2 Matching Anything by Segmenting Anything ‣ 3 Method ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.2.1 </span>MASA Pipeline</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsubsection"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S3.SS2.SSS2" title="In 3.2 Matching Anything by Segmenting Anything ‣ 3 Method ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.2.2 </span>MASA Adapter</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsubsection"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S3.SS2.SSS3" title="In 3.2 Matching Anything by Segmenting Anything ‣ 3 Method ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.2.3 </span>Inference</span></a></li> </ol> </li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S4" title="In Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4 </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/2406.04221v1#S4.SS1" title="In 4 Experiments ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4.1 </span>Experimental Setup</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S4.SS2" title="In 4 Experiments ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4.2 </span>State-of-the-Art Comparison</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S4.SS3" title="In 4 Experiments ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4.3 </span>Ablation Study and Analysis</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S5" title="In Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5 </span>Conclusion</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S6" title="In Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">6 </span>Acknowledgments</span></a></li> <li class="ltx_tocentry ltx_tocentry_part"> <a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#Pt1" title="In Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title">Appendix</span></a> <ol class="ltx_toclist ltx_toclist_part"> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S1a" title="In Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">A </span>Effectiveness on Other Backbones</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S2a" title="In Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">B </span>Zero-shot Evaluation on YoutubeVIS</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S3a" title="In Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">C </span>Visualization of Instance Embeddings</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S4a" title="In Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">D </span>Domain Gap and Adaptation</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S5a" title="In Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">E </span>Impact of Photometric Augmentation</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S6a" title="In Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">F </span>Comparison of Proposal Diversity</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S7" title="In Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">G </span>Compare with Self-Supervised Methods</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S8" title="In Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">H </span>Comparison with VOS-based Methods</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S9" title="In Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">I </span>More Qualitative Results</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/2406.04221v1#S9.SS1" title="In I More Qualitative Results ‣ Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">I.1 </span>Fast Proposal Generation</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S9.SS2" title="In I More Qualitative Results ‣ Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">I.2 </span>Open-Vocabulary Tracking</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S9.SS3" title="In I More Qualitative Results ‣ Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">I.3 </span>Joint Segment and Track Everything</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2406.04221v1#S10" title="In Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">J </span>Implementation Details</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/2406.04221v1#S10.SS1" title="In J Implementation Details ‣ Appendix ‣ Matching Anything by Segmenting Anything"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">J.1 </span>Architecture Detail</span></a></l
