Research map / 3D vision / Face and shape
MarrNet: 3D Shape Reconstruction via 2.5D Sketches
Jiajun Wu, Yifan Wang, Tianfan Xue, Xingyuan Sun, Bill Freeman, Josh Tenenbaum
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 2.5D sketches (depth, normals, silhouette) before 3D shape estimation instead of direct RGB-to-3D prediction | MarrNet · ShapeNet chairs | IoU: 0.52 -> 0.57 | Improved |
| Use MarrNet instead of DRC | 3D reconstruction · PASCAL 3D+ | IoU: 0.34 -> 0.39 | Improved |
| Use MarrNet instead of DRC | 3D reconstruction · PASCAL 3D+ | human preference: 74% | Improved |
| Use MarrNet instead of 3D-VAE-GAN | 3D reconstruction · IKEA | human preference: 61% | Improved |
| Fine-tune on real images while fixing the 3D estimator decoder (ablation) | MarrNet · PASCAL 3D+ | 3D shape quality | Improved |
| Fine-tune on real images without fixing the 3D estimator decoder (ablation) | MarrNet · PASCAL 3D+ | preservation of learned shape prior | Hurt |
Later papers that compared with this method
5 later accepted papers report MarrNet as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- GSIR: Generalizable 3D Shape Interpretation and Reconstruction (ECCV 2020 · better on 2 of 2 shared results)
- Domain-Adaptive Single-View 3D Reconstruction (ICCV 2019 · better on 2 of 2 shared results)
- Pix3D: Dataset and Methods for Single-Image 3D Shape Modeling (CVPR 2018 · better on 3 of 4 shared results)
- Efficient Dense Point Cloud Object Reconstruction using Deformation Vector Fields (ECCV 2018 · better on 2 of 2 shared results)
- Learning to Reconstruct Shapes from Unseen Classes (NeurIPS 2018 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Domain-Adaptive Single-View 3D Reconstruction (ICCV 2019 · similarity 0.88)
- Learning Single-Image 3D Reconstruction by Generative Modelling of Shape, Pose and Shading (IJCV 2019 · similarity 0.88)
- SketchSampler: Sketch-Based 3D Reconstruction via View-Dependent Depth Sampling (ECCV 2022 · similarity 0.87)
- Learning Category-Specific Mesh Reconstruction from Image Collections (ECCV 2018 · similarity 0.87)
- 3D Reconstruction From a Single Sketch via View-Dependent Depth Sampling (TPAMI 2024 · 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.
