Search Tracker: Human-derived Object Tracking In-the-wild Through Large-scale Search And Retrieval

Abstract

Humans use context and scene knowledge to easily localize moving objects in conditions of complex illumination changes, scene clutter and occlusions. In this paper, we present a method to leverage human knowledge in the form of annotated video libraries in a novel search and retrieval based setting to track objects in unseen video sequences. For every video sequence, a document that represents motion information is generated. Documents of the unseen video are queried against the library at multiple scales to find videos with similar motion characteristics. This provides us with coarse localization of objects in the unseen video. We further adapt these retrieved object locations to the new video using an efficient warping scheme. The proposed method is validated on in-the-wild video surveillance datasets where we outperform state-of-the-art appearance-based trackers. We also introduce a new challenging dataset with complex object appearance changes.

Stats

• citations4
• S2 citations—
• github stars0
• HF likes0
• heat score5.24
• arxiv keybency2016search

Related papers

• Introduction Of A Tree-based Technique For Efficient And Real-time Label Retrieval In The Object Tracking System (2022)6.77
• Descriptor Transition Tables For Object Retrieval Using Unconstrained Cluttered Video Acquired Using A Consumer Level Handheld Mobile Device (2016)6.34
• Object Priors For Classifying And Localizing Unseen Actions (2021)9.41
• Object-centric Framework For Video Moment Retrieval (2025)0.00
• Qdtrack: Quasi-dense Similarity Learning For Appearance-only Multiple Object Tracking (2022)15.75
• Use What You Have: Video Retrieval Using Representations From Collaborative Experts (2019)0.00
• Shotfinder: Imagination-driven Open-domain Video Shot Retrieval Via Web Search (2026)0.00
• DMV: Visual Object Tracking Via Part-level Dense Memory And Voting-based Retrieval (2020)0.00