Research map / 3D vision / 3D generation
DeepContext: Context-Encoding Neural Pathways for 3D Holistic Scene Understanding
Yinda Zhang, Mingru Bai, Pushmeet Kohli, Shahram Izadi, Jianxiong Xiao
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 |
|---|---|---|---|
| combine context network with DSS local detector | DSS · SUN RGB-D 3D detection | mAP (%): 43.76 -> 50.50 | Improved |
| add ground-truth alignment instead of estimated transformation (ablation) | DeepContext · SUN RGB-D 3D detection | mAP: +2.19% | Improved |
| add ground-truth template category on top of GT alignment (ablation) | DeepContext · SUN RGB-D 3D detection | mAP: +1.52% | Improved |
| combine context model with DSS on imperfect-template images | DSS · SUN RGB-D (extended set) | mAP (%): 26.80 -> 29.00 | Improved |
| use 3D context network instead of COG for scene understanding precision | DeepContext · SUN RGB-D scene understanding | geometry precision (%): 66.93 -> 71.02 | Improved |
| use transformation network instead of ICP for rotation (sleeping area, with symmetry) | transformation network · SUN RGB-D alignment | rotation accuracy: 96.3% -> 100.0% | Improved |
| use transformation network instead of ICP for rotation (lounging area, no symmetry) | transformation network · SUN RGB-D alignment | rotation accuracy: 58.5% -> 71.7% | Improved |
| use transformation network instead of ICP for translation (sleeping area) | transformation network · SUN RGB-D alignment | translation error (m): 0.473 -> 0.278 | Improved |
| add 3D context network on top of initial floor alignment (sleeping area) | DeepContext · SUN RGB-D layout | floor error mean/median (m): 0.30/0.25 -> 0.10/0.09 | Improved |
| add 3D context network on top of initial wall (office area) | DeepContext · SUN RGB-D layout | wall error mean/median (m): 0.70/0.60 -> 0.60/0.21 | Improved |
| use DeepContext alone instead of DSS | DeepContext · SUN RGB-D 3D detection | speed (s per image): 20 -> 0.5 | Improved |
| use DeepContext alone instead of DSS for chairs | DeepContext · SUN RGB-D 3D detection (chair) | AP: 79.1 -> 44.5 | Hurt |
Closest papers by meaning
- Object-level Scene Context Prediction (TPAMI 2021 · similarity 0.90)
- Semantic Scene Completion From a Single Depth Image (CVPR 2017 · similarity 0.89)
- Holistic 3D Scene Understanding From a Single Image With Implicit Representation (CVPR 2021 · similarity 0.88)
- Lexicon3D: Probing Visual Foundation Models for Complex 3D Scene Understanding (NeurIPS 2024 · similarity 0.87)
- Context and Geometry Aware Voxel Transformer for Semantic Scene Completion (NeurIPS 2024 · similarity 0.87)
- TB-HSU: Hierarchical 3D Scene Understanding with Contextual Affordances (AAAI 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.
