RECORD
SoccerNet-FoulRet: Retrieving Semantically Similar Soccer Foul Videos
Computer Vision and Pattern Recognition · Submitted 7 Oct 2026
abstract compiled · 2026-10-08 · no language-model review
The authors' main claim Semantic foul retrieval remains challenging.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
Refereeing decisions in professional soccer remain inconsistent because referees cannot easily compare a contentious foul against similar past cases. We cast this as a retrieval problem and introduce SoccerNet-FoulRet, the first benchmark for semantic foul retrieval. Given a query foul, the task is to retrieve past fouls judged to be relevant precedents, regardless of camera angle, teams, or appearance. This differs from prior video-to-video retrieval, which matches clips by visual similarity or a shared event. Here, relevance is defined by refereeing interpretation. We build the benchmark from the SoccerNet-MVFoul dataset and evaluate retrieval ability of zero-shot video and vision-language embedders together with a task-specific fine-tuned baseline on 693 human-verified queries and category-relevance labels. Semantic foul retrieval remains challenging. The strongest zero-shot model achieves under 5% HitRate@10 on human-verified precedents, while category-supervised fine-tuning improves category relevance but transfers only modestly to precedent retrieval. We release SoccerNet-FoulRet to establish semantic foul retrieval as an open problem: https://github.com/SoccerNet/sn-foulret.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | Refereeing decisions in professional soccer remain inconsistent because referees cannot easily compare a contentious foul against similar past cases. | background 0.94 | 0.05 | 0.03 | asserted 1.00 |
| s1 | We cast this as a retrieval problem and introduce SoccerNet-FoulRet, the first benchmark for semantic foul retrieval. | contribution 1.00 | 0.93 | 0.08 | asserted 1.00 |
| s2 | Given a query foul, the task is to retrieve past fouls judged to be relevant precedents, regardless of camera angle, teams, or appearance. | contribution 0.57 | 0.23 | 0.04 | asserted 0.99 |
| s3 | This differs from prior video-to-video retrieval, which matches clips by visual similarity or a shared event. | background 0.40 | 0.12 | 0.04 | asserted 1.00 |
| s4 | Here, relevance is defined by refereeing interpretation. | method 0.48 | 0.27 | 0.04 | asserted 1.00 |
| s5 | We build the benchmark from the SoccerNet-MVFoul dataset and evaluate retrieval ability of zero-shot video and vision-language embedders together with a task-specific fine-tuned baseline on 693 human-verified queries and category-relevance labels. | method 0.93 | 0.97 | 0.06 | asserted 0.99 |
| s6 | Semantic foul retrieval remains challenging. | background 0.37 | 0.35 | 0.22 | asserted 1.00 |
| s7 | The strongest zero-shot model achieves under 5% HitRate@10 on human-verified precedents, while category-supervised fine-tuning improves category relevance but transfers only modestly to precedent retrieval. | result 1.00 | 0.93 | 0.63 | asserted 1.00 |
| s8 | We release SoccerNet-FoulRet to establish semantic foul retrieval as an open problem: https://github.com/SoccerNet/sn-foulret. | contribution 0.89 | 0.88 | 0.09 | asserted 0.82 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
Refereeing decisions in professional soccer remain inconsistent because referees cannot easily compare a contentious foul against similar past cases. We cast this as a retrieval problem and introduce SoccerNet-FoulRet, the first benchmark for semantic foul retrieval. Given a query foul, the task is to retrieve past fouls judged to be relevant precedents, regardless of camera angle, teams, or appearance. This differs from prior video-to-video retrieval, which matches clips by visual similarity or a shared event. Here, relevance is defined by refereeing interpretation. We build the benchmark from the SoccerNet-MVFoul dataset and evaluate retrieval ability of zero-shot video and vision-language embedders together with a task-specific fine-tuned baseline on 693 human-verified queries and category-relevance labels. Semantic foul retrieval remains challenging. The strongest zero-shot model achieves under 5% HitRate@10 on human-verified precedents, while category-supervised fine-tuning improves category relevance but transfers only modestly to precedent retrieval. We release SoccerNet-FoulRet to establish semantic foul retrieval as an open problem: https://github.com/SoccerNet/sn-foulret.
Human review
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.