Research map / 3D vision / Face and shape
DiscoNet: Shapes Learning on Disconnected Manifolds for 3D Editing
Eloi Mehr, Ariane Jourdan, Nicolas Thome, Matthieu Cord, Vincent Guitteny
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 pre-learned templates to multi-autoencoder model (ablation) | multiAE · ShapeNet chairs (k=2) | mean distance error (x100): 2.42 -> 1.87 | Improved |
| add pre-learned templates to multi-autoencoder model (ablation) | multiAE · ShapeNet all three categories (k=3) | mean distance error (x100): 1.45 -> 1.25 | Improved |
| add pre-learned templates to multi-autoencoder model (ablation) | multiAE · ShapeNet cars (k=2) | mean distance error (x100): 0.89 -> 0.84 | Improved |
| add pre-learned templates to multi-autoencoder model (ablation) | multiAE · ShapeNet planes (k=2) | mean distance error (x100): 0.95 -> 0.81 | Improved |
| add several autoencoders to single-template model (ablation) | template · ShapeNet all three categories (k=3) | mean distance error (x100): 1.61 -> 1.25 | Improved |
| add several autoencoders to single-template model (ablation) | template · ShapeNet chairs (k=2) | mean distance error (x100): 1.93 -> 1.87 | Improved |
| add several autoencoders to single-template model (ablation) | template · ShapeNet cars (k=2) | mean distance error (x100): 0.86 -> 0.84 | Improved |
| add several autoencoders to single-template model (ablation) | template · ShapeNet planes (k=2) | mean distance error (x100): 0.81 -> 0.81 | No clear effect |
| use DiscoNet instead of FN/AN baseline | FN/AN · ShapeNet chairs (k=2) | mean distance error (x100): 2.22 -> 1.87 | Improved |
| use DiscoNet instead of FN/AN baseline | FN/AN · ShapeNet all three categories (k=3) | mean distance error (x100): 1.46 -> 1.25 | Improved |
| use DiscoNet instead of AtlasNet baseline | AtlasNet · ShapeNet all three categories (k=3) | mean distance error (x100): 1.78 -> 1.25 | Improved |
| add similarity and plausibility energy terms to latent optimization (ablation) | 3D editing optimization · 3D editing (qualitative) | mean distance error (x100) | Improved |
Later papers that compared with this method
1 later accepted paper reports DiscoNet as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- RCDN: Towards Robust Camera-Insensitivity Collaborative Perception via Dynamic Feature-based 3D Neural Modeling (NeurIPS 2024 · better on 3 of 3 shared results)
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
- Manifold Learning in Quotient Spaces (CVPR 2018 · similarity 0.87)
- Variational Autoencoders for Deforming 3D Mesh Models (CVPR 2018 · similarity 0.87)
- Shape Completion Using 3D-Encoder-Predictor CNNs and Shape Synthesis (CVPR 2017 · similarity 0.86)
- ShapeWalk: Compositional Shape Editing Through Language-Guided Chains (CVPR 2024 · similarity 0.85)
- PrEditor3D: Fast and Precise 3D Shape Editing (CVPR 2025 · similarity 0.84)
- Flow-Corrected Shape Optimization: Taming Manifold Drift in High-Dimensional 3D Models (NeurIPS 2026 · similarity 0.84)
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.
