Research map / 3D vision / Face and shape
Shape Inpainting Using 3D Generative Adversarial Network and Recurrent Convolutional Networks
Weiyue Wang, Qiangui Huang, Suya You, Chao Yang, Ulrich Neumann
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 GAN discriminator to the encoder-decoder (3D-ED-GAN instead of VConv-DAE) (ablation) | 3D-ED-GAN · ShapeNet (simulated 3D scanner) | reconstruction error: 7.48% -> 6.55% | Improved |
| add LRCN refinement to 3D-ED-GAN (Hybrid) (ablation) | 3D-ED-GAN · ShapeNet (simulated 3D scanner) | reconstruction error: 6.55% -> 4.74% | Improved |
| add 3D-ED-GAN stage to LRCN (Hybrid) (ablation) | LRCN · ShapeNet (simulated 3D scanner) | reconstruction error: 7.08% -> 4.74% | Improved |
| add unsupervised 3D-ED-GAN pre-training to classifier (Ours-FT) (ablation) | encoder classifier · ModelNet40 | accuracy: 86.1% -> 87.3% | Improved |
| add unsupervised 3D-ED-GAN pre-training to classifier (Ours-FT) (ablation) | encoder classifier · ModelNet10 | accuracy: 90.5% -> 92.6% | Improved |
| use l1 loss instead of l2 loss for LRCN (ablation) | LRCN · ShapeNet | reconstruction quality | Improved |
| use LRCN alone instead of 3D-ED-GAN under increasing random noise (ablation) | LRCN · ShapeNet random noise | reconstruction error | Mixed |
| treat volume as 2D slice sequence with LRCN instead of adding 3D layers | Hybrid network · Model size | parameters / GPU memory: 116.4M -> 26.3M | Improved |
Closest papers by meaning
- Unsupervised 3D Shape Completion Through GAN Inversion (CVPR 2021 · similarity 0.91)
- Inferring 3D Shapes from Image Collections Using Adversarial Networks (IJCV 2020 · similarity 0.90)
- DECOR-GAN: 3D Shape Detailization by Conditional Refinement (CVPR 2021 · similarity 0.89)
- Learning Adversarial 3D Model Generation With 2D Image Enhancer (AAAI 2018 · similarity 0.89)
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 2024 · similarity 0.89)
- Efficient Geometry-Aware 3D Generative Adversarial Networks (CVPR 2022 · similarity 0.88)
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.
