A Lightweight Moment Retrieval System with Global Re-Ranking and Robust Adaptive Bidirectional Temporal Search

2025-04-12

AI Summary

To address the challenges of low efficiency, weak temporal modeling, and incomplete contextual understanding in fine-grained moment retrieval from large-scale video corpora, this paper proposes an interactive video corpus moment retrieval framework. The method introduces three key innovations: (1) a hyper-global re-ranking mechanism that transcends local similarity modeling; (2) robust adaptive bidirectional temporal search (ABTS), which jointly optimizes temporal continuity and computational efficiency; and (3) an integrated pipeline combining keyframe extraction, image hashing-based deduplication, cross-modal similarity modeling, and dynamic temporal pruning. Experimental results demonstrate significant reductions in storage and computational overhead while maintaining high localization accuracy across heterogeneous, multi-source video repositories. The framework exhibits strong scalability and cross-domain robustness, establishing a novel paradigm for large-scale video segment retrieval.

Technology Category

• Computer Vision: Image and Video Retrieval
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs

Abstract

The exponential growth of digital video content has posed critical challenges in moment-level video retrieval, where existing methodologies struggle to efficiently localize specific segments within an expansive video corpus. Current retrieval systems are constrained by computational inefficiencies, temporal context limitations, and the intrinsic complexity of navigating video content. In this paper, we address these limitations through a novel Interactive Video Corpus Moment Retrieval framework that integrates a SuperGlobal Reranking mechanism and Adaptive Bidirectional Temporal Search (ABTS), strategically optimizing query similarity, temporal stability, and computational resources. By preprocessing a large corpus of videos using a keyframe extraction model and deduplication technique through image hashing, our approach provides a scalable solution that significantly reduces storage requirements while maintaining high localization precision across diverse video repositories.

Problem

• Efficiently localize specific segments in large video corpus
• Overcome computational inefficiencies and temporal context limitations
• Reduce storage needs while maintaining high localization precision

Innovation

• SuperGlobal Reranking for query similarity optimization
• Adaptive Bidirectional Temporal Search (ABTS)
• Keyframe extraction and deduplication via image hashing