Research map / 3D vision / 3D generation
Learning Action Maps of Large Environments via First-Person Vision
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Transfer scene and activity information from a source scene to a novel scene using spatial, object, and scene kernels | Action Maps · Novel scenes | F1 | Helped |
| Fit activity observations from all scenes jointly rather than using each scene's observations alone | Action Maps · Office Flr. A | W. Mean F1: 0.69 $\pm$ 0.04 -> 0.73 $\pm$ 0.02 | Helped |
| Add object-detection and scene-classification side information to the spatial kernel (ablation) | Action Maps · Office Flr. A, single scene | W. Mean F1: 0.72 $\pm$ 0.01 -> 0.69 $\pm$ 0.04 | No clear effect |
| Use all side information and jointly fit scenes instead of using only the spatial kernel on one scene (ablation) | Action Maps · Office Flr. A | W. Mean F1: 0.72 $\pm$ 0.01 -> 0.73 $\pm$ 0.02 | Helped |
| Use scene-classification rather than object-detection side information (ablation) | Action Maps · Novel scenes | W. Mean F1: Office Flr. B: 0.67 $\pm$ 0.02 -> 0.69 $\pm$ 0.02; Office Flr. C: 0.55 $\pm$ 0.1 -> 0.56 $\pm$ 0.08; Office Flr. D: 0.51 | Mixed |
| Add object-detection side information to spatial and scene-classification kernels (ablation) | Action Maps · Novel scenes | W. Mean F1: Office Flr. B: 0.69 $\pm$ 0.02 -> 0.54 $\pm$ 0.03; Office Flr. C: 0.56 $\pm$ 0.08 -> 0.63 $\pm$ 0.05; Office Flr. D: 0.5 | Mixed |
| Use detected rather than labelled activity observations (ablation) | Action Maps · Office Flr. A, all scenes | W. Mean F1: 0.73 $\pm$ 0.02 -> 0.71 $\pm$ 0.02 | No clear effect |
| Increase the amount of observed activity demonstrations | Action Maps · Office Flr. A, Office Flr. D, and Home A | F1 | Helped |
Closest papers by meaning
- Learning to Evolve Scenes: Reasoning about Human Activities with Scene Graphs (NeurIPS 2026 · similarity 0.88)
- SAW-Bench: Learning Situated Awareness in the Real World (ICML 2026 · similarity 0.87)
- Grounding Video Models to Actions through Goal Conditioned Exploration (ICLR 2025 · similarity 0.86)
- Action Scene Graphs for Long-Form Understanding of Egocentric Videos (CVPR 2024 · similarity 0.85)
- EgoExoLearn: A Dataset for Bridging Asynchronous Ego- and Exo-centric View of Procedural Activities in Real World (CVPR 2024 · similarity 0.85)
- DIV-FF: Dynamic Image-Video Feature Fields For Environment Understanding in Egocentric Videos (CVPR 2025 · 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.