Research map / 3D vision / NeRF and radiance fields
BallGAN: 3D-aware Image Synthesis with a Spherical Background
Minjung Shin, Yunji Seo, Jeongmin Bae, Young Sun Choi, Hyunsu Kim, Hyeran Byun, Youngjung Uh
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 spherical opaque background to StyleNeRF (BallGAN-S) on CompCars | StyleNeRF · CompCars 256^2 | FID: 8 -> 4.26 | Improved |
| add foreground density loss L_fg (ablation) | BallGAN · FFHQ (batch 16) | FID: 7.87 -> 6.82 | Improved |
| add background transmittance loss L_bg alone (ablation) | BallGAN · FFHQ (batch 16) | FID: 7.87 -> 7.88 | No clear effect |
| add L_bg on top of L_fg (ablation) | BallGAN · FFHQ (batch 16) | FID: 6.82 -> 6.13 | Improved |
| use cube instead of sphere background (ablation) | BallGAN-S · CompCars | convergence | Hurt |
| use sphere background instead of NeRF++ background (ablation) | NeRF foreground, single scene overfitting lf-basket · lf-basket SSO | foreground separation and depth (qualitative) | Improved |
| replace EG3D with BallGAN sphere background | EG3D · FFHQ 512^2 | pose error: 0.007 -> 0.005 | Improved |
| replace EG3D with BallGAN sphere background | EG3D · FFHQ 512^2 | FID: 4.7 -> 5.67 | Hurt |
| replace EG3D with BallGAN sphere background | EG3D · AFHQv2-Cats 512^2 | FID: 2.77 -> 4.72 | Hurt |
| replace EG3D with BallGAN sphere background | EG3D · FFHQ COLMAP | reconstructed points (10^4): 78 -> 79 | No clear effect |
| add sphere background replacing NeRF++ in EpiGRAF (ablation) | EpiGRAF · FFHQ and Cats | foreground separation (qualitative) | Improved |
Later papers that compared with this method
1 later accepted paper reports BallGAN as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- MaTe3D: Mask-Guided Text-based 3D-Aware Portrait Editing (IJCV 2026 · better on 3 of 3 shared results)
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.
- StyleNeRF — StyleNeRF: A Style-based 3D Aware Generator for High-resolution Image Synthesis (ICLR 2022) · better on 6 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
- Improving 3D-aware Image Synthesis with A Geometry-aware Discriminator (NeurIPS 2022 · similarity 0.89)
- 3D-Aware Image Synthesis via Learning Structural and Textural Representations (CVPR 2022 · similarity 0.88)
- SphereHead: Stable 3D Full-head Synthesis with Spherical Tri-plane Representation (ECCV 2024 · similarity 0.88)
- Dual Mapping of 2D StyleGAN for 3D-Aware Image Generation and Manipulation (Student Abstract) (AAAI 2024 · similarity 0.87)
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 2024 · similarity 0.87)
- G-NeRF: Geometry-enhanced Novel View Synthesis from Single-View Images (CVPR 2024 · similarity 0.85)
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.
