Research map / 3D vision / Face and shape
3D Volumetric Modeling with Introspective Neural Networks
Wenlong Huang, Brian Lai, Weijian Xu, Zhuowen Tu
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 |
|---|---|---|---|
| train CNN with introspective reclassification-by-synthesis (3DWINN) instead of plain training (ablation) | volumetric CNN · ModelNet10 FGSM | adversarial error: 27.64% -> 19.05% | Improved |
| 3DWINN vs baseline under iterative FGSM 5 steps (ablation) | volumetric CNN · ModelNet10 | adversarial error: 30.50% -> 19.27% | Improved |
| 3DWINN vs baseline under iterative FGSM 10 steps (ablation) | volumetric CNN · ModelNet10 | adversarial error: 53.19% -> 43.83% | Improved |
| 3DWINN vs baseline clean accuracy (ablation) | volumetric CNN · ModelNet10 supervised | accuracy: 93.1% -> 93.6% | Improved |
| 3DWINN FID vs 3D-DescriptorNet | 3D generative model · ModelNet10 Chair FID | FID: 490 -> 225 | Improved |
| single jointly trained 3DWINN instead of per-category models (ablation) | 3DWINN · ModelNet10 generation | Inception score: 10.25 -> 7.81 | Hurt |
| 3DWINN-separate-multi vs 3D-DescriptorNet | 3D generative model · ModelNet10 generation | Inception score: 11.77 -> 10.25 | Hurt |
| 3DWINN vs 3D-DescriptorNet (unsupervised features + SVM) | 3D generative model · ModelNet10 unsupervised classification | accuracy: 92.4% -> 91.9% | Hurt |
| 3DWINN vs ORION supervised | 3D classifier · ModelNet10 supervised | accuracy: 93.8% -> 93.6% | Hurt |
Closest papers by meaning
- Disentangled3D: Learning a 3D Generative Model With Disentangled Geometry and Appearance From Monocular Images (CVPR 2022 · similarity 0.86)
- Wasserstein Introspective Neural Networks (CVPR 2018 · similarity 0.86)
- Learning Adversarial 3D Model Generation With 2D Image Enhancer (AAAI 2018 · similarity 0.85)
- Volumetric and Multi-View CNNs for Object Classification on 3D Data (CVPR 2016 · similarity 0.85)
- Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling (NeurIPS 2016 · similarity 0.85)
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 2024 · similarity 0.83)
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.
