Research map / 3D vision / 3D generation
View Independent Generative Adversarial Network for Novel View Synthesis
Xiaogang Xu, Ying-Cong Chen, Jiaya Jia
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 view-independent loss (ablation) | VI-GAN · ShapeNet chair | L1 error: 16.43 -> 12.56 | Improved |
| add view-independent loss (ablation) | VI-GAN · ShapeNet chair | SSIM: 0.82 -> 0.87 | Improved |
| add pose prediction loss (ablation) | VI-GAN · ShapeNet chair | SSIM: 0.80 -> 0.87 | Improved |
| add pose prediction loss (ablation) | VI-GAN · ShapeNet chair | L1 error: 16.81 -> 12.56 | Improved |
| add CoordConv in generator (ablation) | VI-GAN · ShapeNet chair | L1 error: 14.35 -> 12.56 | Improved |
| add CoordConv in generator (ablation) | VI-GAN · ShapeNet chair | SSIM: 0.85 -> 0.87 | Improved |
| add AdaIN conditioned on target pose in decoder (ablation) | VI-GAN · ShapeNet chair | L1 error: 14.02 -> 12.56 | Improved |
| add AdaIN conditioned on target pose in decoder (ablation) | VI-GAN · ShapeNet chair | SSIM: 0.84 -> 0.87 | Improved |
| pretrain encoder with VI-GAN view-independent constraint | 3D landmark network · AFLW2000-3D | NME (%): 12.7% -> 6.8% | Improved |
| perturb input camera pose within about 10 degrees | VI-GAN encoder · Multi-PIE | mean L1 error | No clear effect |
| increase target-source pose difference beyond 90 degrees | VI-GAN decoder · Multi-PIE | mean L1 error | Hurt |
| VI-GAN instead of CR-GAN | Multi-PIE face rotation · Multi-PIE (aligned) | FID: 8.76 -> 6.51 | Improved |
Closest papers by meaning
- A Style-Based GAN Encoder for High Fidelity Reconstruction of Images and Videos (ECCV 2022 · similarity 0.87)
- Towards Unsupervised Learning of Generative Models for 3D Controllable Image Synthesis (CVPR 2020 · similarity 0.87)
- High-Fidelity 3D GAN Inversion by Pseudo-Multi-View Optimization (CVPR 2023 · similarity 0.87)
- Efficient Geometry-Aware 3D Generative Adversarial Networks (CVPR 2022 · similarity 0.87)
- Dual Encoder GAN Inversion for High-Fidelity 3D Head Reconstruction from Single Images (NeurIPS 2024 · similarity 0.87)
- Multi-View Consistent Generative Adversarial Networks for 3D-Aware Image Synthesis (CVPR 2022 · 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.
