Research map / Detection and segmentation / Open-set and few-shot detection
Self-Supervised Object Detection from Egocentric Videos
Peri Akiva, Jing Huang, Kevin J Liang, Rama Kovvuri, Xingyu Chen, Matt Feiszli, Kristin Dana, Tal Hassner
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 |
|---|---|---|---|
| train DEVI on egocentric Ego4D frames instead of exocentric YouTubeBB-8M frames (ablation) | DEVI · Ego4D val | AP50: +3.86% AP50, +1.42% AR1, +8.80% AR10, +16.17% AR100 | Improved |
| replace Object Residual Module with K-Means clustering on raw features (ablation) | DEVI · Ego4D val | AP50 | Hurt |
| add multi-view loss with the Object Residual Module (ablation) | DEVI · Ego4D val | AP50: +1.62% AP50 over baseline | Improved |
| add scale-regression loss on top of multi-view loss and ORM (ablation) | DEVI · Ego4D val | AP50: final AP50 6.51 | Improved |
| use DEVI end-to-end single-stage training instead of multi-stage FreeSOLO | class-agnostic detection · Ego4D val | AR100: 17.00 -> 22.03 with about 72h -> 36h training | Improved |
| DEVI versus FreeSOLO on lower-complexity EgoObjects | DEVI · EgoObjects val | AP50 and AR1: AP50 14.96 vs 15.70; AR1 6.47 vs 8.30; AR100 39.43 vs 32.90 | Mixed |
| apply DEVI trained on Ego4D to COCO without any COCO training | DEVI · COCO val | AP50: AP50 8.03 vs FreeSOLO 12.20; AR100 25.93 vs 15.30 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FreeSOLO — FreeSOLO: Learning To Segment Objects Without Annotations (CVPR 2022) · better on 5 of 7 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
- EgoObjects: A Large-Scale Egocentric Dataset for Fine-Grained Object Understanding (ICCV 2023 · similarity 0.94)
- Towards Stable Self-Supervised Object Representations in Unconstrained Egocentric Video (CVPR 2026 · similarity 0.91)
- EgoTracks: A Long-term Egocentric Visual Object Tracking Dataset (NeurIPS 2023 · similarity 0.90)
- Ego-Only: Egocentric Action Detection without Exocentric Transferring (ICCV 2023 · similarity 0.89)
- SOS! Self-Supervised Learning over Sets of Handled Objects in Egocentric Action Recognition (ECCV 2022 · similarity 0.89)
- Instance Tracking in 3D Scenes from Egocentric Videos (CVPR 2024 · similarity 0.88)
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.
