Temporal Feature Distillation for Label-Efficient Precise Event Spotting in Sports Videos

2026-07-12

AI Summary

This work addresses the challenge of distinguishing visually similar yet semantically distinct adjacent frames in sports videos and the tendency of existing self-distillation methods to overlook critical motion cues. To this end, the authors propose a semi-supervised temporal feature distillation framework based on Vision Transformers. Innovatively performing temporal alignment at the feature level—rather than at the projection head output—the approach incorporates a supervised warm-up strategy and a multi-scale Transformer gated shift module to enhance sensitivity to event boundaries. Evaluated on four fine-grained sports video datasets, the method significantly outperforms current fully and semi-supervised approaches: it achieves a 4.54-point mAP gain with only 10% labeled data and matches or exceeds the performance of 100% fully supervised training on two datasets using just 80% labeled data.

Abstract

Precise Event Spotting (PES) requires distinguishing visually similar yet semantically distinct adjacent frames, making it fundamentally different from image classification and coarse action recognition. Although self-distillation methods such as DINO have shown strong representation learning ability in images, we find that directly applying them to PES is ineffective: without supervised guidance, subtle but crucial motion cues are often suppressed as noise, leading to representations that are insensitive to precise event boundaries. To address this, we propose Temporal Feature Distillation, a semi-supervised objective that aligns temporally informative backbone features, rather than projection-head outputs, to preserve motion-sensitive and boundary-aware cues for frame-level localization. A supervised warm-up with a ramp-up schedule further stabilizes training by ensuring that meaningful event cues are learned before unlabeled distillation begins. We also introduce Transformer Gate Shift, a multi-scale gated shifting module that injects motion-aware temporal information into Vision Transformers. Experiments on four fine-grained sports benchmarks show consistent improvements over fully supervised and semi-supervised baselines. Under 10\% supervision on FSPerf, our method improves mAP by 4.54 points over the strongest competing approach, and with only 80\% labeled data, it matches or surpasses the fully supervised 100\% baseline on two of the four datasets.

Problem

• Precise Event Spotting
• Temporal Feature Distillation
• Label-Efficient Learning
• Sports Video Analysis
• Frame-level Localization

Innovation

• Temporal Feature Distillation
• Precise Event Spotting
• Semi-supervised Learning
• Transformer Gate Shift
• Motion-aware Representation