Research map / 3D vision / 3D generation
What Do Single-View 3D Reconstruction Networks Learn?
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 image-based shape retrieval instead of encoder-decoder reconstruction | Single-view 3D reconstruction on ShapeNet · ShapeNet (55 classes) | mIoU: 0.47 -> 0.57 | Helped |
| Use image classification to choose a cluster mean shape instead of encoder-decoder reconstruction | Single-view 3D reconstruction on ShapeNet · ShapeNet (55 classes) | mIoU: 0.38 -> 0.46; 0.46 -> 0.46; 0.47 -> 0.46 | Mixed |
| Use the Oracle nearest training shape as the prediction | ShapeNet single-view reconstruction · ShapeNet (55 classes) | mIoU: 0.57 -> 0.76 | Helped |
| Evaluate in a viewer-centered rather than object-centered coordinate frame | OGN and Matryoshka Networks · ShapeNet | mIoU | Hurt |
| Evaluate retrieval in a viewer-centered rather than object-centered coordinate frame | Retrieval · ShapeNet | mIoU | Hurt |
| Use viewer-centered reconstruction and assess surface accuracy with F-score at a 1% threshold | Retrieval · ShapeNet (viewer-centered) | F-score (@1%) | Mixed |
| Use viewer-centered reconstruction and assess surface accuracy with F-score at a 1% threshold | AtlasNet · ShapeNet (viewer-centered) | F-score (@1%) | Mixed |
| Train and evaluate on ShapeNet's canonical object-centered, within-class shape split | Single-view 3D reconstruction methods · ShapeNet | mIoU: 0.47 -> 0.76 | Helped |
Closest papers by meaning
- Perspective Transformer Nets: Learning Single-View 3D Object Reconstruction without 3D Supervision (NeurIPS 2016 · similarity 0.91)
- Know Your Neighbors: Improving Single-View Reconstruction via Spatial Vision-Language Reasoning (CVPR 2024 · similarity 0.87)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.87)
- Less Decoder is More Encoder: Geometric Representation Learning from Novel View Synthesis (NeurIPS 2026 · similarity 0.87)
- LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias (ICLR 2025 · similarity 0.87)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.86)
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.