Research map / 3D vision / Face and shape
Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling
Read the paperCheck 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 discriminator features learned by 3D-GAN for classification rather than VConv-DAE features | Unsupervised 3D shape classification · ModelNet40 | classification accuracy: 75.5% -> 83.3% | Helped |
| Use discriminator features learned by 3D-GAN for classification rather than VConv-DAE features | Unsupervised 3D shape classification · ModelNet10 | classification accuracy: 80.5% -> 91.0% | Helped |
| Use 3D-GAN rather than the strongest reported supervised baseline | 3D shape classification · ModelNet40 | classification accuracy: 91.4% -> 83.3% | Hurt |
| Use 3D-GAN rather than the strongest reported baseline | 3D shape classification · ModelNet10 | classification accuracy: 93.8% -> 91.0% | Hurt |
| Train separate 3D-VAE-GAN models per category instead of one model jointly across categories (ablation) | 3D-VAE-GAN · IKEA | mean average precision: 45.2 -> 53.1 | Helped |
| Train separate 3D-VAE-GAN models per category instead of one model jointly across categories (ablation) | 3D-VAE-GAN · IKEA categories | average precision: 49.1 -> 63.2; 31.9 -> 46.3; 42.6 -> 47.2; 34.8 -> 40.7; 79.8 -> 78.8; 33.1 -> 42.3 | Mixed |
| Use adaptive discriminator updates, updating only when prior-batch accuracy is not higher than 80% | 3D-GAN training · 3D object generation | generation quality | Helped |
Closest papers by meaning
- Learning Adversarial 3D Model Generation With 2D Image Enhancer (AAAI 2018 · similarity 0.91)
- Rank3DGAN: Semantic Mesh Generation Using Relative Attributes (AAAI 2020 · similarity 0.89)
- Synthesizing 3D Shapes From Silhouette Image Collections Using Multi-Projection Generative Adversarial Networks (CVPR 2019 · similarity 0.89)
- Visual Object Networks: Image Generation with Disentangled 3D Representations (NeurIPS 2018 · similarity 0.89)
- 3D generation on ImageNet (ICLR 2023 · similarity 0.89)
- Disentangled3D: Learning a 3D Generative Model With Disentangled Geometry and Appearance From Monocular Images (CVPR 2022 · similarity 0.89)
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.