Vision-Language Models Show Promise in Judging Olympic Diving

Researchers have explored the potential of vision-language models (VLMs) for assessing the quality of Olympic diving performances. A proposed framework leverages VLMs' semantic reasoning and phase-level sub-scores, combined with TF-IDF vectorization and ensemble learning, to predict final competition scores. While standalone VLMs showed limited correlation, the ensemble approach achieved a Spearman correlation of 0.67, indicating that VLM-generated explanations are valuable for sports performance evaluation. AI

影响 VLMs show potential as assistive tools for explainable and semi-automated sports performance evaluation.

排序理由 Academic paper detailing a new methodology for action quality assessment using VLMs. [lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

• AQA-7 benchmark dataset
• GitHub
• Olympic diving
• vision-language model

• 论文
• 产品

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →