Research map / 3D vision / 3D generation
EgoLoc: Revisiting 3D Object Localization from Egocentric Videos with Visual Queries
Jinjie Mai, Abdullah Hamdi, Silvio Giancola, Chen Zhao, Bernard Ghanem
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 |
|---|---|---|---|
| replace Matterport re-localization with COLMAP egocentric SfM for camera poses (ablation) | Ego4D VQ3D baseline · Ego4D VQ3D test | Success rate (%): 8.71 -> 77.27 | Improved |
| use detection peaks instead of last tracking response (ablation) | EgoLoc · Ego4D VQ3D validation | Success rate (%): 73.78 -> 78.05 | Improved |
| use multiple detection peaks instead of last peak (ablation) | EgoLoc · Ego4D VQ3D validation | Success rate (%): 78.05 -> 80.49 | Improved |
| weight multi-view 3D predictions by detection confidence instead of simple mean (ablation) | EgoLoc · Ego4D VQ3D validation | Success rate (%): 79.88 -> 80.49 | Improved |
| weighted multi-view aggregation on test server (ablation) | EgoLoc · Ego4D VQ3D test | Success rate (%): 86.36 -> 87.12 | Improved |
| use ground-truth 2D tracking instead of detected peaks (ablation) | EgoLoc · Ego4D VQ3D validation | Success* (%): 97.39 -> 98.04/98.69 | No clear effect |
| replace DPT monocular depth with multi-view triangulation (ablation) | EgoLoc · Ego4D VQ3D validation | Success rate (%): 80.49 -> 56.1 | Hurt |
| replace DPT depth with random depth (ablation) | EgoLoc · Ego4D VQ3D validation | Success rate (%): 80.49 -> 34.15 | Hurt |
| add Gaussian noise to DPT depth scale and shift (ablation) | EgoLoc · Ego4D VQ3D validation | Success rate (%): 80.49 -> 78.66 | Hurt |
Later papers that compared with this method
2 later accepted papers report EgoLoc-v1 as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- EAGLE: Episodic Appearance- and Geometry-aware Memory for Unified 2D-3D Visual Query Localization in Egocentric Vision (AAAI 2026 · better on 4 of 4 shared results)
- Embodied VideoAgent: Persistent Memory from Egocentric Videos and Embodied Sensors Enables Dynamic Scene Understanding (ICCV 2025 · better on 2 of 4 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- RELOCATE: A Simple Training-Free Baseline for Visual Query Localization Using Region-Based Representations (CVPR 2025 · similarity 0.89)
- GPVK-VL: Geometry-Preserving Virtual Keyframes for Visual Localization under Large Viewpoint Changes (CVPR 2025 · similarity 0.86)
- Spatial Reasoning with Vision-Language Models in Ego-Centric Multi-View Scenes (ICLR 2026 · similarity 0.86)
- Instance Tracking in 3D Scenes from Egocentric Videos (CVPR 2024 · similarity 0.86)
- Layered Motion Fusion: Lifting Motion Segmentation to 3D in Egocentric Videos (CVPR 2025 · similarity 0.86)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.86)
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.
