Research map / Video / Action recognition
SOS! Self-Supervised Learning over Sets of Handled Objects in Egocentric Action Recognition
Victor Escorcia, Ricardo Guerrero, Xiatian Zhu, Brais Martinez
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 SOS pre-training to OIC without LT loss (ablation) | OIC · EPIC-KITCHENS-100 val (long-tail) | overall action accuracy: 8.7 -> 10.8 | Improved |
| add SOS pre-training to OIC without LT loss (ablation) | OIC · EPIC-KITCHENS-100 val (long-tail) | overall noun accuracy: 21.1 -> 25.0 | Improved |
| add LT loss on top of SOS pre-training (ablation) | OIC · EPIC-KITCHENS-100 val (long-tail) | overall action accuracy: 10.8 -> 9.4 | Mixed |
| add LT loss on top of SOS pre-training (ablation) | OIC · EPIC-KITCHENS-100 val (long-tail) | overall verb accuracy: 20.8 -> 30.3 | Improved |
| replace supervised ImageNet init with SwAV self-supervised init (ablation) | OIC · EPIC-KITCHENS-100 val | noun/action accuracy: 44.5 -> 45.8 noun; 27.4 -> 27.9 action | Improved |
| add on-domain SwAV adaptation on object regions (ablation) | OIC · EPIC-KITCHENS-100 val | noun accuracy: 45.8 -> 47.0 | Improved |
| replace SwAV with set-based SwAV-S in on-domain pre-training (ablation) | OIC · EPIC-KITCHENS-100 val | action accuracy: 28.7 -> 30.2 | Improved |
| fuse OIC with TSM (visual) | TSM · EPIC-KITCHENS-100 val | action accuracy: 35.6 -> 39.0 | Improved |
| fuse OIC with X-ViT using test-time augmentation | X-ViT · EPIC-KITCHENS-100 val | action accuracy: 44.3 -> 45.7 | Improved |
| fuse OIC with TSM vs ensembling two TSM models | TSM · EPIC-KITCHENS-100 val | noun accuracy: 48.3 -> 52.3 | Improved |
| fuse OIC with TSM on unseen participants verb | TSM · EPIC-KITCHENS-100 val (unseen participants) | verb accuracy: 54.6 -> 53.1 | Hurt |
| fuse OIC with TSM | TSM · EGTEA | mean class accuracy: 60.7 -> 61.3 | Improved |
Closest papers by meaning
- Interactive Prototype Learning for Egocentric Action Recognition (ICCV 2021 · similarity 0.89)
- Self-Supervised Object Detection from Egocentric Videos (ICCV 2023 · similarity 0.89)
- Self-supervised visual learning from interactions with objects (ECCV 2024 · similarity 0.88)
- Modeling Fine-Grained Hand-Object Dynamics for Egocentric Video Representation Learning (ICLR 2025 · similarity 0.88)
- First Person Action Recognition Using Deep Learned Descriptors (CVPR 2016 · similarity 0.88)
- Egocentric Action Recognition by Automatic Relation Modeling (TPAMI 2022 · 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.
