Research map / Detection and segmentation / Open-set and few-shot detection
A Comparative Analysis and Study of Multiview CNN Models for Joint Object Categorization and Pose Estimation
Read the paperCheck 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 |
|---|---|---|---|
| Early branching at Pool5 (EBM) instead of late branching (LBM) (ablation) | joint category and pose CNN · RGBD Split 1 | Pose accuracy % (<22.5°): 71.29 vs 68.25 | Helped |
| Cross-product joint softmax (CPM) instead of separate category and pose outputs (ablation) | joint category and pose CNN · RGBD Split 1 | Category accuracy %: 80.68 vs 89.63 (PM) | Hurt |
| Cross-product joint softmax (CPM) instead of separate category and pose outputs (ablation) | joint category and pose CNN · Pascal3D+ Pascal12 test | Pose accuracy % (<22.5): 49.39 vs 47.34 (PM) | Mixed |
| Early branching at Pool5 with pose subnetwork (EBM) vs parallel independent networks (PM) (ablation) | joint category and pose CNN · Pascal3D+ Pascal12 test | Pose accuracy % (<22.5): 51.37 vs 47.34 (PM) | Helped |
| Shrinking the pose subnetwork size in EBM (4096 to 100 units) (ablation) | EBM · RGBD Split 1 | Pose accuracy % (<22.5°): 71.49 (4096) vs 67.70 (200) | Mixed |
| Using EBM (800) instead of prior state of the art | joint category and pose estimation · RGBD Split 2 | Pose accuracy % (AAAI): 79.30 vs 77.31 | Helped |
| Adding ImageNet images to training for joint EBM model (ablation) | EBM (800) · Pascal3D+ Pascal12 + ImageNet test | Pose accuracy % (<22.5): 67.26 vs 51.89 (ImageNet test) | Helped |
| Pretraining FC6/FC7 of pose branch with category weights (PM) vs random init (EBM) (ablation) | pose branch · RGBD Split 1 | Pose accuracy % (<22.5°): 69.58 (PM) vs 71.29 (EBM 800) | Hurt |
Closest papers by meaning
- RotationNet: Joint Object Categorization and Pose Estimation Using Multiviews From Unsupervised Viewpoints (CVPR 2018 · similarity 0.92)
- Dual Pose-invariant Embeddings: Learning Category and Object-specific Discriminative Representations for Recognition and Retrieval (CVPR 2024 · similarity 0.90)
- Multi-Path Learning for Object Pose Estimation Across Domains (CVPR 2020 · similarity 0.90)
- Convolutional Pose Machines (CVPR 2016 · similarity 0.90)
- Deep Co-Occurrence Feature Learning for Visual Object Recognition (CVPR 2017 · similarity 0.88)
- NOPE: Novel Object Pose Estimation from a Single Image (CVPR 2024 · similarity 0.87)
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.