Research map / 3D vision / Face and shape
To The Point: Correspondence-driven monocular 3D category reconstruction
Filippos Kokkinos, Iasonas Kokkinos
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 |
|---|---|---|---|
| Use correspondence-driven pose and deformation optimization instead of the strongest reported baseline | TTP on CUB with keypoint supervision · CUB | PCK: 85.5 -> 93.4 | Improved |
| Use correspondence-driven reconstruction without keypoint supervision instead of ACSM with 8 camera hypotheses | TTP on CUB · CUB | mIoU: 0.693 -> 0.752 | Improved |
| Add keypoint supervision | TTP · PASCAL3D+ car | 3D mIoU: 0.665 -> 0.67 | No clear effect |
| Add texture loss (ablation) | TTP with keypoint supervision · CUB | mIoU: 0.752 -> 0.765 | Improved |
| Add texture loss (ablation) | TTP without keypoint supervision · CUB | mIoU: 0.667 -> 0.749 | Improved |
| Add visibility loss (ablation) | TTP without keypoint supervision · CUB | PCK: 9.3 -> 50.9 | Improved |
| Add equivariance loss (ablation) | TTP without keypoint supervision · CUB | PCK: 28.2 -> 50.9 | Improved |
| Use a deformable basis instead of a rigid template (ablation) | TTP with keypoint supervision · CUB | mIoU: 0.657 -> 0.765 | Improved |
| Use a deformable basis instead of a rigid template (ablation) | TTP without keypoint supervision · CUB | mIoU: 0.646 -> 0.749 | Improved |
| Increase basis size from 16 to 64 components (ablation) | TTP with keypoint supervision · CUB | mIoU: 0.765 -> 0.775 | Improved |
| Increase basis size from 16 to 64 components (ablation) | TTP without keypoint supervision · CUB | no clear effect: 0.749 -> 0.752 | No clear effect |
| Increase pose/deformation optimization iterations from 1 to 4 (ablation) | TTP with keypoints · CUB | mIoU: 0.732 -> 0.758 | Improved |
Closest papers by meaning
- Learning Monocular 3D Reconstruction of Articulated Categories From Motion (CVPR 2021 · similarity 0.90)
- Unsupervised 3D Reconstruction and Grouping of Rigid and Non-Rigid Categories (TPAMI 2020 · similarity 0.88)
- LASR: Learning Articulated Shape Reconstruction From a Monocular Video (CVPR 2021 · similarity 0.88)
- Shape-Pose Ambiguity in Learning 3D Reconstruction from Images (AAAI 2021 · similarity 0.88)
- Learning Category-Specific Mesh Reconstruction from Image Collections (ECCV 2018 · similarity 0.88)
- Unsupervised Learning of Category-Level 3D Pose from Object-Centric Videos (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.
