Research map / 3D vision / 3D generation
RGBD-GAN: Unsupervised 3D Representation Learning From Natural Image Datasets via RGBD Image Synthesis
Atsuhiro Noguchi, Tatsuya Harada
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 camera conditioning and the RGBD 3D consistency loss (ablation) | PGGAN · FFHQ | FID: 28.5 -> 30.3 | Hurt |
| Add camera conditioning and the RGBD 3D consistency loss (ablation) | PGGAN · ShapeNet Car | FID: 16.7 -> 14.5 | Improved |
| Add camera conditioning and the RGBD 3D consistency loss (ablation) | StyleGAN · FFHQ | FID: 20.9 -> 24.2 | Hurt |
| Add camera conditioning and the RGBD 3D consistency loss (ablation) | StyleGAN · ShapeNet Car | FID: 15.5 -> 13.5 | Improved |
| Add the proposed 3D consistency loss (ablation) | DeepVoxels · FFHQ | V_depth: 0.00487 -> 0.00067 | Improved |
| Add the proposed 3D consistency loss (ablation) | DeepVoxels · FFHQ | V_color: 0.0153 -> 0.0072 | Improved |
| Add the proposed 3D consistency loss (ablation) | DeepVoxels · ShapeNet Car | V_depth: 0.00045 -> 0.00030 | Improved |
| Add the proposed 3D consistency loss (ablation) | DeepVoxels · ShapeNet Car | V_color: 0.0398 -> 0.0283 | Improved |
| Add the proposed 3D consistency loss (ablation) | DeepVoxels · FFHQ | FID: 19.4 -> 21.1 | Hurt |
| Add the proposed 3D consistency loss (ablation) | DeepVoxels · ShapeNet Car | FID: 28.6 -> 31.2 | Hurt |
| Use occlusion-aware projection rather than weighted-sum depth projection (ablation) | DeepVoxels-based model · ShapeNet Car | 360° rotation consistency | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DeepVoxels — DeepVoxels: Learning Persistent 3D Feature Embeddings (CVPR 2019) · better on 5 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- HoloGAN: Unsupervised Learning of 3D Representations From Natural Images (ICCV 2019 · similarity 0.89)
- 3D generation on ImageNet (ICLR 2023 · similarity 0.89)
- Progressive Learning of 3D Reconstruction Network From 2D GAN Data (TPAMI 2023 · similarity 0.88)
- Efficient Geometry-Aware 3D Generative Adversarial Networks (CVPR 2022 · similarity 0.88)
- Improving 3D-aware Image Synthesis with A Geometry-aware Discriminator (NeurIPS 2022 · similarity 0.88)
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 2024 · similarity 0.87)
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.
