Research map / 3D vision / 3D generation
Pre-Train, Self-Train, Distill: A Simple Recipe for Supersizing 3D Reconstruction
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 |
|---|---|---|---|
| Replace category-specific models with the unified model distilled from experts | 3D reconstruction · PASCAL3D+ | IoU: 0.42 -> 0.48 | Helped |
| Replace category-specific models with the unified model distilled from experts | 3D reconstruction · Co3D | Chamfer distance: 0.182 -> 0.176 | Helped |
| Increase the number of distilled category-specific experts from over 75 to over 125 (ablation) | Unified reconstruction model · Co3D unseen classes | Chamfer loss: 0.308 -> 0.239 | Helped |
| Distill category-specific experts into a unified model trained across over 75 categories | 3D reconstruction · Co3D unseen classes | Chamfer loss: 0.368 -> 0.308 | Helped |
Closest papers by meaning
- Unsupervised Learning of 3D Object Categories From Videos in the Wild (CVPR 2021 · similarity 0.90)
- Learning Monocular 3D Reconstruction of Articulated Categories From Motion (CVPR 2021 · similarity 0.89)
- Multiview Compressive Coding for 3D Reconstruction (CVPR 2023 · similarity 0.88)
- What Do Single-View 3D Reconstruction Networks Learn? (CVPR 2019 · similarity 0.88)
- MESC-3D:Mining Effective Semantic Cues for 3D Reconstruction from a Single Image (CVPR 2025 · similarity 0.88)
- RefineAny3D: Depth Refinement as Semantic Alignment for Monocular 3D Detection (NeurIPS 2026 · 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.