Research map / 3D vision / Face and shape
3D Interpreter Networks for Viewer-Centered Wireframe Modeling
Jiajun Wu, Tianfan Xue, Joseph J. Lim, Yuandong Tian, Joshua B. Tenenbaum, Antonio B. Torralba, William T. Freeman
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 keypoint refinement (fine-level network) to the keypoint estimator (ablation) | 3D-INN · FLIC | PCK: about 2% for a normalized distance >= 0.15 | Improved |
| add synthetic-data pre-training stages instead of end-to-end training on real images only (ablation) | 3D-INN · IKEA chairs | average recall (structure): about 20% higher | Improved |
| add synthetic-data pre-training stages instead of end-to-end training on real images only (ablation) | 3D-INN · IKEA chairs | average recall (viewpoint): about 40% higher | Improved |
| add fine-tuning with the projection layer on real 2D images (ablation) | 3D-INN · IKEA | accuracy at RMSE threshold 0.15: about 5% | Improved |
| use learned 3D interpreter instead of optimization-based reprojection minimization | 3D-INN · IKEA | RMSE-recall | Improved |
| use 3D-INN instead of prior viewpoint estimation method | 3D-INN · IKEA viewpoint | average recall: about 40% (relative) | Improved |
| use 3D-INN instead of Mdshift for keypoints | 3D-INN · CUB-200-2011 | PCP (%) / average error: 69.1 -> 66.7 PCP; 1.39 -> 1.36 AE | Mixed |
| use 3D-INN instead of retrained prior method for swivel chair keypoints | 3D-INN · Keypoint-5 swivel chair | PCP (%): 69.2 -> 78.5 | Improved |
| use 3D-INN instead of V&K for viewpoint with different training data | 3D-INN · PASCAL 3D+ sofa | AVP: 43.8 -> 45.8 | Improved |
Closest papers by meaning
- Single Image 3D Interpreter Network (ECCV 2016 · similarity 0.97)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.88)
- 2D Features Are All You Need for 3D Shape Understanding (ECCV 2026 · similarity 0.88)
- 3D-RCNN: Instance-Level 3D Object Reconstruction via Render-and-Compare (CVPR 2018 · similarity 0.87)
- Structured3D: A Large Photo-realistic Dataset for Structured 3D Modeling (ECCV 2020 · similarity 0.87)
- CoReNet: Coherent 3D Scene Reconstruction from a Single RGB Image (ECCV 2020 · 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.
