Research map / Video / Video-language understanding
PRVQL: Progressive Knowledge-guided Refinement for Robust Egocentric Visual Query Localization
Bing Fan, Yunhe Feng, Yapeng Tian, James Chenhao Liang, Yuewei Lin, Yan Huang, Heng Fan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add appearance knowledge generation (AKG) (ablation) | PRVQL · Ego4D val | stAP25: 0.23 -> 0.26 | Improved |
| add spatial knowledge generation (SKG) (ablation) | PRVQL · Ego4D val | stAP25: 0.23 -> 0.24 | Improved |
| combine AKG and SKG (ablation) | PRVQL · Ego4D val | stAP25: 0.23 -> 0.27 | Improved |
| increase stages from 1 to 3 (ablation) | PRVQL · Ego4D val | success: 55.37 -> 57.93 | Improved |
| increase stages from 3 to 4 (ablation) | PRVQL · Ego4D val | tAP25: 0.35 -> 0.33 | Hurt |
| use cross-attention instead of addition to combine knowledge into query (ablation) | PRVQL QFR · Ego4D val | tAP25: 0.31 -> 0.35 | Improved |
| use cross-attention instead of concatenation in QFR (ablation) | PRVQL QFR · Ego4D val | stAP25: 0.25 -> 0.27 | Improved |
| set confidence threshold tau to 0.7 rather than 0.6 or 0.8 (ablation) | AKG · Ego4D val | tAP25: 0.32 -> 0.35 | Improved |
| increase number of AKG boxes from 3 to 4 (ablation) | AKG · Ego4D val | tAP25: 0.35 -> 0.36 | Mixed |
| use PRVQL versus VQLoC on small objects | PRVQL · Ego4D val small objects | tAP25: 0.047 (VQLoC) vs 0.036 | Hurt |
Closest papers by meaning
- Single-Stage Visual Query Localization in Egocentric Videos (NeurIPS 2023 · similarity 0.94)
- RELOCATE: A Simple Training-Free Baseline for Visual Query Localization Using Region-Based Representations (CVPR 2025 · similarity 0.90)
- EgoThinker: Unveiling Egocentric Reasoning with Spatio-Temporal CoT (NeurIPS 2025 · similarity 0.86)
- Localizing Events in Videos with Multimodal Queries (CVPR 2025 · similarity 0.85)
- GPVK-VL: Geometry-Preserving Virtual Keyframes for Visual Localization under Large Viewpoint Changes (CVPR 2025 · similarity 0.85)
- Referencing Where to Focus: Improving Visual Grounding with Referential Query (NeurIPS 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
