Research map / 3D vision / NeRF and radiance fields
3D-Aware Multi-Class Image-to-Image Translation With NeRFs
Senmao Li, Joost van de Weijer, Yaxing Wang, Fahad Shahbaz Khan, Meiqin Liu, Jian Yang
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 |
|---|---|---|---|
| Use 3D-aware multi-class translation instead of StarGANv2 | CelebA-HQ · CelebA-HQ | TC: 10.250 -> 3.743 | Improved |
| Use 3D-aware multi-class translation instead of StarGANv2 | AFHQ · AFHQ | TC: 3.025 -> 2.067 | Improved |
| Add the U-net-like adaptor to the initialized model (ablation) | 3D-aware I2I translation · AFHQ | TC: 2.612 -> 2.324 | Improved |
| Add the hierarchical representation constraint (ablation) | 3D-aware I2I translation · AFHQ | FID: 23.1 -> 16.1 | Improved |
| Add the relative regularization loss (ablation) | 3D-aware I2I translation · AFHQ | TC: 2.204 -> 2.067 | Improved |
| Add the relative regularization loss (ablation) | 3D-aware I2I translation · AFHQ | FID: 16.1 -> 15.3 | Improved |
| Initialize multi-class StyleNeRF from an unconditional StyleNeRF instead of training the conditional model from scratch (ablation) | Multi-class StyleNeRF · AFHQ and CelebA-HQ | image quality | Improved |
| Use one mapping network with concatenated class embedding and noise (ablation) | Conditional 3D-aware generator · 3D-aware multi-class generation | target-specific image realism | Hurt |
| Use two mapping networks without concatenating their outputs (ablation) | Conditional 3D-aware generator · 3D-aware multi-class generation | 3D-aware generation | Hurt |
| Predict 3D location and viewing direction instead of predicting the feature map (ablation) | 3D-aware I2I adaptor · Image translation | image quality | Hurt |
| Use StarGANv2 for viewpoint-conditioned translation | StarGANv2 · CelebA-HQ | vLPIPS: 0.032 vs 0.101 | Hurt |
Closest papers by meaning
- Efficient Geometry-Aware 3D Generative Adversarial Networks (CVPR 2022 · similarity 0.89)
- DeepI2I: Enabling Deep Hierarchical Image-to-Image Translation by Transferring from GANs (NeurIPS 2020 · similarity 0.88)
- Benchmarking and Analyzing 3D-aware Image Synthesis with a Modularized Codebase (NeurIPS 2023 · similarity 0.88)
- Improving 3D-aware Image Synthesis with A Geometry-aware Discriminator (NeurIPS 2022 · similarity 0.88)
- 3D-Aware Image Synthesis via Learning Structural and Textural Representations (CVPR 2022 · similarity 0.88)
- Reference-Based 3D-Aware Image Editing with Triplanes (CVPR 2025 · 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.
