{"id":{"repo_id":"vilnius","oai_identifier":"oai:vu.lt:elaba:308337913"},"canonical_url":"https://search.ndltd.org/etd/vilnius/oai:vu.lt:elaba:308337913","repository":{"repo_id":"vilnius","name":"Vilnius University","base_url":"https://epublications.vu.lt/oai"},"display":{"title":"Fechtavimo judesių klasifikavimas pagal kūno ir špagos taškų sekas taikant mokymąsi iš keleto pavyzdžių /","abstract":"This paper presents a fencing movement classification system based on few­shot learning. The system integrates fencer body keypoint tracking, sword tip detection, and a Spatial Temporal Graph Convolutional Network combined with Prototypical Networks and Relation Networks architectures, enabling classification of movement classes and additional attributes, arm position and target zone, using only standard camera footage. A dataset of 130 annotated video clips from world­level competitions was collected, covering 7 movement classes. A sword tip tracking architecture was reconstructed using the ”YOLOv8­seg” model (mAP50 = 0.858). Two body keypoint extraction methods were compared: ”MediaPipe Pose Landmarker” and ”YOLOv8­pose”. Models were trained using episodic training with unseen test classes. Best test accuracy across all models: Relation Network with ”YOLOv8­pose” ­71.5% (5­shot), Prototypical Network with ”MediaPipe” ­66.4% (5­shot). Arm position attribute was recognised reliably (76-­90% depending on the model), target zone reached &#8776;40­-43% due to frequent tip detection failures. The Prototypical Network with ”MediaPipe” showed the most stable generalisation behaviour, while the Relation Network with ”YOLOv8­pose” achieved the highest test accuracy ­consistently improving with increasing shot count.","abstract_html":"This paper presents a fencing movement classification system based on few­shot learning. The system integrates fencer body keypoint tracking, sword tip detection, and a Spatial Temporal Graph Convolutional Network combined with Prototypical Networks and Relation Networks architectures, enabling classification of movement classes and additional attributes, arm position and target zone, using only standard camera footage. A dataset of 130 annotated video clips from world­level competitions was collected, covering 7 movement classes. A sword tip tracking architecture was reconstructed using the ”YOLOv8­seg” model (mAP50 = 0.858). Two body keypoint extraction methods were compared: ”MediaPipe Pose Landmarker” and ”YOLOv8­pose”. Models were trained using episodic training with unseen test classes. Best test accuracy across all models: Relation Network with ”YOLOv8­pose” ­71.5% (5­shot), Prototypical Network with ”MediaPipe” ­66.4% (5­shot). Arm position attribute was recognised reliably (76-­90% depending on the model), target zone reached &amp;#8776;40­-43% due to frequent tip detection failures. The Prototypical Network with ”MediaPipe” showed the most stable generalisation behaviour, while the Relation Network with ”YOLOv8­pose” achieved the highest test accuracy ­consistently improving with increasing shot count.","abstract_has_math":false,"creators":["Malinauskas, Daumantas,"],"institution":"Institutional Repository of Vilnius University","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026","date_published":"2026","updated_at":"2026-11-19T10:12:59Z","subjects":[],"languages":["lit"],"rights":["info:eu-repo/semantics/openAccess"],"rights_urls":[],"identifier_entries":[],"person_identifier_entries":[]},"links":{"outbound_url":"https://repository.vu.lt/VU:ELABAETD308337913&prefLang=en_US","outbound_label":"Repository record","outbound_source":"dc:identifier","record_url":"https://repository.vu.lt/VU:ELABAETD308337913&prefLang=en_US","record_label":"Repository record","record_kind":"record","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":"open","embargo_until":null,"access_note":null},"source_record":{"url":"https://epublications.vu.lt/oai?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Avu.lt%3Aelaba%3A308337913","prefix":"oai_dc"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["Malinauskas, Daumantas,"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2026"]},{"key":"dc:publisher","label":"Institution","values":["Institutional Repository of Vilnius University"]},{"key":"dc:relation","label":"Dc Relation","values":["https://epublications.vu.lt/object/elaba:308337913/308337913.pdf"]},{"key":"dc:type","label":"Dc Type","values":["info:eu-repo/semantics/bachelorThesis"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["lit"]},{"key":"dc:rights","label":"Dc Rights","values":["info:eu-repo/semantics/openAccess"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://repository.vu.lt/VU:ELABAETD308337913&prefLang=en_US"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["This paper presents a fencing movement classification system based on few­shot learning. The system integrates fencer body keypoint tracking, sword tip detection, and a Spatial Temporal Graph Convolutional Network combined with Prototypical Networks and Relation Networks architectures, enabling classification of movement classes and additional attributes, arm position and target zone, using only standard camera footage. A dataset of 130 annotated video clips from world­level competitions was collected, covering 7 movement classes. A sword tip tracking architecture was reconstructed using the ”YOLOv8­seg” model (mAP50 = 0.858). Two body keypoint extraction methods were compared: ”MediaPipe Pose Landmarker” and ”YOLOv8­pose”. Models were trained using episodic training with unseen test classes. Best test accuracy across all models: Relation Network with ”YOLOv8­pose” ­71.5% (5­shot), Prototypical Network with ”MediaPipe” ­66.4% (5­shot). Arm position attribute was recognised reliably (76-­90% depending on the model), target zone reached &#8776;40­-43% due to frequent tip detection failures. The Prototypical Network with ”MediaPipe” showed the most stable generalisation behaviour, while the Relation Network with ”YOLOv8­pose” achieved the highest test accuracy ­consistently improving with increasing shot count."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Fechtavimo judesių klasifikavimas pagal kūno ir špagos taškų sekas taikant mokymąsi iš keleto pavyzdžių /","Fencing movement classification based on body and épée point sequences using few-shot learning."]}]}],"canonical_facts":{"dc:creator":["Malinauskas, Daumantas,"],"dc:date":["2026"],"dc:description":["This paper presents a fencing movement classification system based on few­shot learning. The system integrates fencer body keypoint tracking, sword tip detection, and a Spatial Temporal Graph Convolutional Network combined with Prototypical Networks and Relation Networks architectures, enabling classification of movement classes and additional attributes, arm position and target zone, using only standard camera footage. A dataset of 130 annotated video clips from world­level competitions was collected, covering 7 movement classes. A sword tip tracking architecture was reconstructed using the ”YOLOv8­seg” model (mAP50 = 0.858). Two body keypoint extraction methods were compared: ”MediaPipe Pose Landmarker” and ”YOLOv8­pose”. Models were trained using episodic training with unseen test classes. Best test accuracy across all models: Relation Network with ”YOLOv8­pose” ­71.5% (5­shot), Prototypical Network with ”MediaPipe” ­66.4% (5­shot). Arm position attribute was recognised reliably (76-­90% depending on the model), target zone reached &#8776;40­-43% due to frequent tip detection failures. The Prototypical Network with ”MediaPipe” showed the most stable generalisation behaviour, while the Relation Network with ”YOLOv8­pose” achieved the highest test accuracy ­consistently improving with increasing shot count."],"dc:format":["application/pdf"],"dc:identifier":["https://repository.vu.lt/VU:ELABAETD308337913&prefLang=en_US"],"dc:language":["lit"],"dc:publisher":["Institutional Repository of Vilnius University"],"dc:relation":["https://epublications.vu.lt/object/elaba:308337913/308337913.pdf"],"dc:rights":["info:eu-repo/semantics/openAccess"],"dc:title":["Fechtavimo judesių klasifikavimas pagal kūno ir špagos taškų sekas taikant mokymąsi iš keleto pavyzdžių /","Fencing movement classification based on body and épée point sequences using few-shot learning."],"dc:type":["info:eu-repo/semantics/bachelorThesis"]},"updated_at":"2026-11-19T10:12:59Z"}