FineBadminton: A Multi-Level Dataset for Fine-Grained Badminton Video Understanding

2025-08-10

AI Summary

Fine-grained understanding of high-speed sports such as badminton remains challenging due to inherent spatiotemporal complexity and a critical scarcity of high-quality, domain-specific annotated data. Method: This work introduces FBBench—the first benchmark for badminton video understanding—built upon a novel multi-level annotation dataset covering fundamental actions, tactical semantics, and decision-level evaluations. We propose a human-in-the-loop annotation pipeline integrating multimodal large language models (MLLMs) for candidate generation and expert refinement, augmented by stroke-centered keyframe selection and coordinate-guided visual compression. Contribution/Results: Comprehensive evaluation on FBBench reveals severe limitations of current MLLMs in spatiotemporal reasoning and tactical analysis. Our approach significantly improves performance across all tiers, establishing FBBench as a reproducible benchmark, delivering high-quality labeled resources, and providing an effective technical baseline for fine-grained sports intelligence.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Video Understanding & Activity Analysis
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Fine-grained analysis of complex and high-speed sports like badminton presents a significant challenge for Multimodal Large Language Models (MLLMs), despite their notable advancements in general video understanding. This difficulty arises primarily from the scarcity of datasets with sufficiently rich and domain-specific annotations. To bridge this gap, we introduce FineBadminton, a novel and large-scale dataset featuring a unique multi-level semantic annotation hierarchy (Foundational Actions, Tactical Semantics, and Decision Evaluation) for comprehensive badminton understanding. The construction of FineBadminton is powered by an innovative annotation pipeline that synergistically combines MLLM-generated proposals with human refinement. We also present FBBench, a challenging benchmark derived from FineBadminton, to rigorously evaluate MLLMs on nuanced spatio-temporal reasoning and tactical comprehension. Together, FineBadminton and FBBench provide a crucial ecosystem to catalyze research in fine-grained video understanding and advance the development of MLLMs in sports intelligence. Furthermore, we propose an optimized baseline approach incorporating Hit-Centric Keyframe Selection to focus on pivotal moments and Coordinate-Guided Condensation to distill salient visual information. The results on FBBench reveal that while current MLLMs still face significant challenges in deep sports video analysis, our proposed strategies nonetheless achieve substantial performance gains. The project homepage is available at https://finebadminton.github.io/FineBadminton/.

Problem

• Addresses lack of domain-specific datasets for badminton video analysis
• Introduces multi-level semantic annotations for comprehensive sports understanding
• Proposes benchmark to evaluate MLLMs on tactical reasoning

Innovation

• Multi-level semantic annotation hierarchy for badminton
• MLLM-generated proposals with human refinement pipeline
• Hit-Centric Keyframe Selection and Coordinate-Guided Condensation