Research map / Architectures and efficiency / Neural architecture search
ProbRes: Probabilistic Jump Diffusion for Open-World Egocentric Activity Recognition
Sanjoy Kundu, Shanmukha Vellamcheti, Sathyanarayanan N. Aakur
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 ProbRes search to LaVila | LaVila · GTEA Gaze (L1) | Activity: 29.02 -> 33.80 | Improved |
| add ProbRes search to EgoVLP | EgoVLP · GTEA Gaze (L1) | Activity: 17.07 -> 19.12 | Improved |
| add ProbRes to LaVila on EK100 | LaVila · EK100 (L1) | Activity: 18.81 -> 18.89 | Improved |
| add ProbRes to LaVila on CharadesEgo | LaVila · CharadesEgo (L1) | Activity: 11.34 -> 13.71 | Improved |
| add ProbRes to LaVila with LLM-generated kitchen search space | LaVila · GTEA Gaze (L2) | Activity WUPS: 38.34 -> 43.28 | Improved |
| add ProbRes to LaVila with broad LLM search space | LaVila · GTEA Gaze (L3) | Activity WUPS: 46.06 -> 47.71 | Improved |
| add ProbRes to EgoVLP at L2 | EgoVLP · GTEA Gaze (L2) | Object WUPS: 56.77 -> 56.27 | Hurt |
| add ProbRes to LaVila at L2 | LaVila · GTEA Gaze+ (L2) | Action WUPS: 30.36 -> 28.03 | Hurt |
| use knowledge-graph neuro-symbolic KGL instead of VLM | KGL · GTEA Gaze (L1) | WUPS: 51.31 -> 35.83 | Hurt |
| increase search iterations (ablation) | ProbRes (LaVila, GTEA Gaze L3) · GTEA Gaze (L3) | WUPS / exact match | Mixed |
| add ConceptNet priors to exploration (lambda about 0.6 vs 0) (ablation) | ProbRes · GTEA Gaze (L3) | WUPS / exact match | Improved |
| add ConceptNet knowledge priors (ablation) | ProbRes · GTEA Gaze (L3) | performance | Improved |
Closest papers by meaning
- Discovering Novel Actions from Open World Egocentric Videos with Object-Grounded Visual Commonsense Reasoning (ECCV 2024 · similarity 0.87)
- Egocentric Activity Recognition and Localization on a 3D Map (ECCV 2022 · similarity 0.86)
- Learning Action Maps of Large Environments via First-Person Vision (CVPR 2016 · similarity 0.86)
- Interactive Prototype Learning for Egocentric Action Recognition (ICCV 2021 · similarity 0.86)
- Action Scene Graphs for Long-Form Understanding of Egocentric Videos (CVPR 2024 · similarity 0.85)
- Ego-Exo4D: Understanding Skilled Human Activity from First- and Third-Person Perspectives (CVPR 2024 · similarity 0.84)
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.
