Research map / 3D vision / Depth and stereo
Pairwise Decomposition of Image Sequences for Active Multi-View Recognition
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 |
|---|---|---|---|
| use all view pairs rather than only the top M weighted pairs (ablation) | Pairwise recognition · ModelNet10, random 6-view sequences | accuracy: 90.6 -> 91.9 | Helped |
| use all view pairs rather than only the top M weighted pairs (ablation) | Pairwise recognition · ModelNet40, random 6-view sequences | accuracy: 88.8 -> 89.5 | Helped |
| weight pair contributions by learned relative-pose confidence (ablation) | Pairwise recognition with all pairs · ModelNet10, random 6-view sequences | accuracy: 91.2 -> 91.9 | Helped |
| weight pair contributions by learned relative-pose confidence (ablation) | Pairwise recognition with all pairs · ModelNet40, random 6-view sequences | accuracy: 89.0 -> 89.5 | Helped |
| add depth images to greyscale input (ablation) | Pairwise recognition · ModelNet10, constrained average | accuracy: 90.9 -> 91.3 | Helped |
| add depth images to greyscale input (ablation) | Pairwise recognition · ModelNet40, constrained average | accuracy: 88.6 -> 89.2 | Helped |
| use a straight path instead of randomly selected adjacent moves (ablation) | Pairwise recognition · ModelNet10, unconstrained average | accuracy: 88.5 -> 90.9 | Helped |
| use a straight path instead of randomly selected adjacent moves (ablation) | Pairwise recognition · ModelNet40, unconstrained average | accuracy: 88.2 -> 89.0 | Helped |
| allow next-best-view selection over the full viewing sphere instead of adjacent views only (ablation) | Active recognition · ModelNet40, active recognition average | accuracy: 89.4 -> 90.7 | Helped |
| optimise the trajectory using all images along it rather than following next-best-view predictions (ablation) | Active recognition · ModelNet40, active recognition average | accuracy: 89.4 -> 89.8 | Helped |
| use greyscale + depth instead of greyscale alone (ablation) | Pairwise recognition · ModelNet10, constrained 3-view | accuracy: 88.5 -> 88.8 | Helped |
| use greyscale + depth instead of greyscale alone (ablation) | Pairwise recognition · ModelNet40, constrained 12-view | accuracy: 90.7 -> 91.1 | Helped |
Closest papers by meaning
- Multi-Path Learning for Object Pose Estimation Across Domains (CVPR 2020 · similarity 0.88)
- A Comparative Analysis and Study of Multiview CNN Models for Joint Object Categorization and Pose Estimation (ICML 2016 · similarity 0.87)
- RotationNet: Joint Object Categorization and Pose Estimation Using Multiviews From Unsupervised Viewpoints (CVPR 2018 · similarity 0.87)
- DVANet: Disentangling View and Action Features for Multi-View Action Recognition (AAAI 2024 · similarity 0.85)
- Dual Pose-invariant Embeddings: Learning Category and Object-specific Discriminative Representations for Recognition and Retrieval (CVPR 2024 · similarity 0.84)
- Multiview Scene Graph (NeurIPS 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.