Research map / 3D vision / 3D generation
Towards Unsupervised Learning of Generative Models for 3D Controllable Image Synthesis
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 unpaired background-image supervision (ablation) | Ours · Indoor | FID: 120 -> 88 | Helped |
| add unpaired background-image supervision (ablation) | Ours · Car | FID: 65 -> 44 | Helped |
| replace 3D primitives with a 2D feature generator (ablation) | Ours · Car | FID: 44 -> 80 | Hurt |
| replace 3D primitives with a 2D feature generator (ablation) | Ours · Indoor | FID: 88 -> 107 | Hurt |
| use sphere instead of point-cloud primitives (ablation) | the multi-primitive model · Car without background | FID: 38 -> 33 | Helped |
| use sphere instead of cuboid primitives (ablation) | the multi-primitive model · Car without background | FID: 38 -> 33 | Helped |
| remove the 2D generator and directly train a deformable 3D primitive (ablation) | the multi-primitive model · Car without background | FID: 33 -> 69 | Hurt |
| represent the whole scene with one primitive instead of separate object primitives (ablation) | the multi-primitive model · Car without background | FID_t: 45 -> 38 | Helped |
| represent the whole scene with one primitive instead of separate object primitives (ablation) | the multi-primitive model · Car without background | FID_R: 45 -> 44 | Helped |
| add geometric consistency loss (ablation) | the multi-primitive model · Car without background | performance | No clear effect |
| use one primitive for the entire scene instead of object-wise primitives (ablation) | the multi-primitive model · camera viewpoint changes | object identity consistency | Hurt |
Closest papers by meaning
- GRAF: Generative Radiance Fields for 3D-Aware Image Synthesis (NeurIPS 2020 · similarity 0.91)
- Visual Object Networks: Image Generation with Disentangled 3D Representations (NeurIPS 2018 · similarity 0.90)
- Learning 3D-Aware Image Synthesis With Unknown Pose Distribution (CVPR 2023 · similarity 0.90)
- Multi-View Consistent Generative Adversarial Networks for 3D-Aware Image Synthesis (CVPR 2022 · similarity 0.90)
- Improving 3D-aware Image Synthesis with A Geometry-aware Discriminator (NeurIPS 2022 · similarity 0.90)
- CAD: Photorealistic 3D Generation via Adversarial Distillation (CVPR 2024 · similarity 0.89)
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.