{"id":431,"date":"2026-04-01T10:14:17","date_gmt":"2026-04-01T09:14:17","guid":{"rendered":"https:\/\/vision.inf.uniri.hr\/?page_id=431"},"modified":"2026-04-01T10:21:46","modified_gmt":"2026-04-01T09:21:46","slug":"sportsight-project-info","status":"publish","type":"page","link":"https:\/\/vision.inf.uniri.hr\/?page_id=431","title":{"rendered":"SportSight – Project info"},"content":{"rendered":"\n

Code: <\/strong>uniri-iz-25-103
Acronym:<\/strong> SportSight
Project Title: <\/strong>Computer Vision System for Athlete Tracking and Movement Analysis
Project Title (hr):<\/strong> Sustav ra\u010dunalnog vida za pra\u0107enje sporta\u0161a i analizu pokreta
Project manager: <\/strong>Miran Pobar
Project starts<\/strong>: 01. 10. 2025.
Duration:<\/strong> 48 Months<\/p>\n\n\n\n

Project abstract: <\/strong><\/h4>\n\n\n\n

The project’s goal is to develop computer vision models for tracking athletes in video, analyzing movements, and automatically comparing the performance of actions against a reference. The main challenge is the continuous tracking of players in recordings due to frequent occlusions and abrupt changes in direction. Comparing actions is made difficult by the variable relationship between the camera and the athlete, differences in the dynamics and execution style of actions, athlete physiology, and the reliability of methods for video-based pose detection.<\/p>\n\n\n\n

Starting with state-of-the-art multi-object tracking algorithms, we will enhance player tracking to be more resilient to occlusions and incorporate movement models adapted for sports (e.g., for sudden changes in direction). To train more robust player tracking models, we will generate targeted synthetic data using a game engine or generative artificial intelligence techniques. The synthetic data will focus on complex situations involving occlusions and rapid movement, with known ground truth regarding player identity and position, which, when using real recordings, requires significant time and costs associated with manual annotation.<\/p>\n\n\n\n

The development of these models will enable the creation of a system capable of providing coaches and players with useful feedback and insights for technical improvement.<\/p>\n\n\n\n

Project abstract(hr):<\/strong><\/h4>\n\n\n\n

Cilj projekta je razviti modele ra\u010dunalnog vida za pra\u0107enje sporta\u0161a na videu, analizu pokreta te automatizirano uspore\u0111ivanje izvedbe akcija s referentnom. Glavni izazov je neprekidno pra\u0107enje igra\u010da na snimci zbog \u010destog prekrivanja i naglih promjena smjera. Usporedbu akcija ote\u017eava varijabilan odnos kamere i sporta\u0161a, razlike u dinamici i na\u010dinu izvo\u0111enja akcija te fizionomiji sporta\u0161a te pouzdanost metoda za detekciju poze iz videa.<\/p>\n\n\n\n

Po\u010dev\u0161i od suvremenih algoritama za pra\u0107enje vi\u0161e objekata, unaprijedit \u0107emo pra\u0107enje igra\u010da koje \u0107e biti otpornije na prekrivanja te uklju\u010diti modele kretanja prilago\u0111ene za sport (npr. za nagle promjene smjera).<\/p>\n\n\n\n

Kako bismo nau\u010dili robusnije modele za pra\u0107enje igra\u010da, generirat \u0107emo ciljane sinteti\u010dke podatke kori\u0161tenjem game enginea ili postupcima generativne umjetne inteligencije. Sinteti\u010dki podaci bit \u0107e fokusirani na kompleksne situacije s prekrivanjem i brzim kretanjem, uz poznato \u010dinjeni\u010dno stanje o identitetu i poziciji igra\u010da, \u0161to u slu\u010daju kori\u0161tenja stvarnih snimaka zahtijeva zna\u010dajno vrijeme i tro\u0161kove povezane s ru\u010dnim ozna\u010davanjem.<\/p>\n\n\n\n

Razvojem ovih modela omogu\u0107it \u0107e se razvoj sustava koji mo\u017ee trenerima i igra\u010dima pru\u017eati korisne povratne informacije i uvide u tehni\u010dka pobolj\u0161anja.<\/p>\n\n\n\n

Team Members: <\/strong><\/p>\n\n\n\n

prof. dr. sc. Marina Iva\u0161i\u0107-Kos<\/p>\n\n\n\n

Doctoral Students: <\/strong><\/p>\n\n\n\n

Kristina Host
Ivona Frankovi\u0107<\/p>\n","protected":false},"excerpt":{"rendered":"

Code: uniri-iz-25-103 Acronym: SportSight Project Title: Computer Vision System for Athlete Tracking and Movement Analysis Project Title (hr): Sustav ra\u010dunalnog vida za pra\u0107enje sporta\u0161a i analizu pokreta Project manager: Miran Pobar Project starts: 01. 10. 2025. Duration: 48 Months Project abstract: The project’s goal is to develop computer vision models for tracking athletes in video, […]<\/p>\n","protected":false},"author":2,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"wp-custom-template-old-projects","meta":{"footnotes":""},"class_list":["post-431","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/vision.inf.uniri.hr\/index.php?rest_route=\/wp\/v2\/pages\/431","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/vision.inf.uniri.hr\/index.php?rest_route=\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/vision.inf.uniri.hr\/index.php?rest_route=\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/vision.inf.uniri.hr\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/vision.inf.uniri.hr\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=431"}],"version-history":[{"count":3,"href":"https:\/\/vision.inf.uniri.hr\/index.php?rest_route=\/wp\/v2\/pages\/431\/revisions"}],"predecessor-version":[{"id":437,"href":"https:\/\/vision.inf.uniri.hr\/index.php?rest_route=\/wp\/v2\/pages\/431\/revisions\/437"}],"wp:attachment":[{"href":"https:\/\/vision.inf.uniri.hr\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=431"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}