{"id":{"repo_id":"uwo","oai_identifier":"oai:uwo.scholaris.ca:20.500.14721/32671"},"canonical_url":"https://search.dev.ndltd.org/etd/uwo/oai:uwo.scholaris.ca:20.500.14721/32671","repository":{"repo_id":"uwo","name":"Western University","base_url":"https://uwo.scholaris.ca/server/oai/request"},"display":{"title":"Dynamically Instance-Guided Adaptation: A Backward-free Approach for Test-Time Domain Adaptive Semantic Segmentation","abstract":"Semantic segmentation models often fail when deployed in new target domains due to domain shifts. Test-Time Domain Adaptation for Semantic Segmentation (TTDA-Seg) aims to adapt models efficiently during inference without target labels, but existing methods struggle with efficiency (requiring backward optimization) or effectiveness (often inadequately addressing semantic shifts), and can suffer from error accumulation. This thesis proposes Dynamically Instance-Guided Adaptation (DIGA), a novel, efficient, and backward-free approach for TTDA-Seg. DIGA leverages information from each test instance to guide its adaptation without requiring backpropagation, thus avoiding costly optimization and error accumulation. The method consists of two main components: a Distribution Adaptation Module (DAM) that adaptively combines source and instance Batch Normalization (BN) statistics for robust feature representation, and a Semantic Adaptation Module (SAM) which constructs a dynamic non-parametric classifier using historical and instance-aware prototypes to refine semantic predictions. Evaluated on five diverse benchmark datasets, DIGA outperforms state-of-the-art TTDA-Seg methods in segmentation accuracy while maintaining high computational efficiency. This work presents DIGA as a robust and practical approach for Test-Time Domain Adaptation in semantic segmentation, highlighting the effectiveness of instance-guided adaptation for handling domain shifts during inference.","abstract_html":"Semantic segmentation models often fail when deployed in new target domains due to domain shifts. Test-Time Domain Adaptation for Semantic Segmentation (TTDA-Seg) aims to adapt models efficiently during inference without target labels, but existing methods struggle with efficiency (requiring backward optimization) or effectiveness (often inadequately addressing semantic shifts), and can suffer from error accumulation. This thesis proposes Dynamically Instance-Guided Adaptation (DIGA), a novel, efficient, and backward-free approach for TTDA-Seg. DIGA leverages information from each test instance to guide its adaptation without requiring backpropagation, thus avoiding costly optimization and error accumulation. The method consists of two main components: a Distribution Adaptation Module (DAM) that adaptively combines source and instance Batch Normalization (BN) statistics for robust feature representation, and a Semantic Adaptation Module (SAM) which constructs a dynamic non-parametric classifier using historical and instance-aware prototypes to refine semantic predictions. Evaluated on five diverse benchmark datasets, DIGA outperforms state-of-the-art TTDA-Seg methods in segmentation accuracy while maintaining high computational efficiency. This work presents DIGA as a robust and practical approach for Test-Time Domain Adaptation in semantic segmentation, highlighting the effectiveness of instance-guided adaptation for handling domain shifts during inference.","abstract_has_math":false,"creators":["Wang, Wei"],"institution":"The University of Western Ontario","degree_name":"M Sc","degree_level":null,"degree_discipline":"Computer Science","degree_department":null,"school":null,"contributors":[],"advisors":["Wang, Boyu","Charles X. Ling"],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025-06-12","date_published":"2025-06-12","updated_at":"2026-07-27T21:55:58Z","subjects":["Test-Time Domain Adaptation","Semantic Segmentation","Domain Shift","Instance-Guided Adaptation","Non-Parametric Adaptation","Backward-Free Adaptation"],"languages":["en_ca"],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/20.500.14721/32671","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Wang, Boyu","Charles X. Ling"]},{"key":"dc:creator","label":"Author","values":["Wang, Wei"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2025-07-10T19:34:42Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2025-07-10T19:34:42Z"]},{"key":"dc:date.issued","label":"Date","values":["2025-06-12"]},{"key":"dc:publisher","label":"Institution","values":["The University of Western Ontario"]},{"key":"dc:type","label":"Dc Type","values":["thesis"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Computer Science"]},{"key":"thesis:degree_name","label":"Degree Name","values":["M Sc"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Test-Time Domain Adaptation","Semantic Segmentation","Domain Shift","Instance-Guided Adaptation","Non-Parametric Adaptation","Backward-Free Adaptation"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en_ca"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/20.500.14721/32671"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["The thesis cover page in the PDF document includes references to Western University’s previous institutional repository platform, known as Scholarship@Western, and links to that platform (beginning with ir.lib.uwo.ca). In citing or referring to this thesis, use the DOI or handle from this page instead. Sample citation: Author name, \"Thesis title.\" (Year). Western University Open Repository. https://doi.org/10.71858/123456."]},{"key":"dc:description.abstract","label":"Abstract","values":["Semantic segmentation models often fail when deployed in new target domains due to domain shifts. Test-Time Domain Adaptation for Semantic Segmentation (TTDA-Seg) aims to adapt models efficiently during inference without target labels, but existing methods struggle with efficiency (requiring backward optimization) or effectiveness (often inadequately addressing semantic shifts), and can suffer from error accumulation. This thesis proposes Dynamically Instance-Guided Adaptation (DIGA), a novel, efficient, and backward-free approach for TTDA-Seg. DIGA leverages information from each test instance to guide its adaptation without requiring backpropagation, thus avoiding costly optimization and error accumulation. The method consists of two main components: a Distribution Adaptation Module (DAM) that adaptively combines source and instance Batch Normalization (BN) statistics for robust feature representation, and a Semantic Adaptation Module (SAM) which constructs a dynamic non-parametric classifier using historical and instance-aware prototypes to refine semantic predictions. Evaluated on five diverse benchmark datasets, DIGA outperforms state-of-the-art TTDA-Seg methods in segmentation accuracy while maintaining high computational efficiency. This work presents DIGA as a robust and practical approach for Test-Time Domain Adaptation in semantic segmentation, highlighting the effectiveness of instance-guided adaptation for handling domain shifts during inference."]},{"key":"dc:title","label":"Title","values":["Dynamically Instance-Guided Adaptation: A Backward-free Approach for Test-Time Domain Adaptive Semantic Segmentation"]}]}],"canonical_facts":{"dc:contributor.advisor":["Wang, Boyu","Charles X. Ling"],"dc:creator":["Wang, Wei"],"dc:date.accessioned":["2025-07-10T19:34:42Z"],"dc:date.available":["2025-07-10T19:34:42Z"],"dc:date.issued":["2025-06-12"],"dc:description":["The thesis cover page in the PDF document includes references to Western University’s previous institutional repository platform, known as Scholarship@Western, and links to that platform (beginning with ir.lib.uwo.ca). In citing or referring to this thesis, use the DOI or handle from this page instead. Sample citation: Author name, \"Thesis title.\" (Year). Western University Open Repository. https://doi.org/10.71858/123456."],"dc:description.abstract":["Semantic segmentation models often fail when deployed in new target domains due to domain shifts. Test-Time Domain Adaptation for Semantic Segmentation (TTDA-Seg) aims to adapt models efficiently during inference without target labels, but existing methods struggle with efficiency (requiring backward optimization) or effectiveness (often inadequately addressing semantic shifts), and can suffer from error accumulation. This thesis proposes Dynamically Instance-Guided Adaptation (DIGA), a novel, efficient, and backward-free approach for TTDA-Seg. DIGA leverages information from each test instance to guide its adaptation without requiring backpropagation, thus avoiding costly optimization and error accumulation. The method consists of two main components: a Distribution Adaptation Module (DAM) that adaptively combines source and instance Batch Normalization (BN) statistics for robust feature representation, and a Semantic Adaptation Module (SAM) which constructs a dynamic non-parametric classifier using historical and instance-aware prototypes to refine semantic predictions. Evaluated on five diverse benchmark datasets, DIGA outperforms state-of-the-art TTDA-Seg methods in segmentation accuracy while maintaining high computational efficiency. This work presents DIGA as a robust and practical approach for Test-Time Domain Adaptation in semantic segmentation, highlighting the effectiveness of instance-guided adaptation for handling domain shifts during inference."],"dc:identifier.uri":["https://hdl.handle.net/20.500.14721/32671"],"dc:language.iso":["en_ca"],"dc:publisher":["The University of Western Ontario"],"dc:subject":["Test-Time Domain Adaptation","Semantic Segmentation","Domain Shift","Instance-Guided Adaptation","Non-Parametric Adaptation","Backward-Free Adaptation"],"dc:title":["Dynamically Instance-Guided Adaptation: A Backward-free Approach for Test-Time Domain Adaptive Semantic Segmentation"],"dc:type":["thesis"],"thesis:degree_discipline":["Computer Science"],"thesis:degree_name":["M Sc"]},"updated_at":"2026-07-27T21:55:58Z"}