Research map / 3D vision / NeRF and radiance fields
GRAM-HD: 3D-Consistent Image Generation at High Resolution with Generative Radiance Manifolds
Jianfeng Xiang, Jiaolong Yang, Yu Deng, Xin Tong
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 |
|---|---|---|---|
| do super-resolution on radiance manifolds with 2D CNN instead of GRAM direct rendering | GRAM · FFHQ 256 | FID: 15.0 -> 11.8 | Improved |
| use manifold super-resolution at 1024^2 | GRAM · FFHQ 1024 inference | running time: 6.69s -> 0.36s | Improved |
| 3D-consistent manifold super-resolution vs image-space upsampling methods | GRAM-HD vs StyleNeRF/StyleSDF/EG3D · FFHQ 1024 FID | FID: 9.45 -> 12.0 | Hurt |
| 3D-consistent manifold super-resolution vs image-space upsampling methods | GRAM-HD vs StyleNeRF · FFHQ 1024 NeuS reconstruction | PSNR: 30.0 -> 33.8 | Improved |
| add cross-resolution consistency loss L_cons (ablation) | GRAM-HD · FFHQ 256 | FID: 16.9 -> 11.8 | Improved |
| add patch adversarial loss L_patch (ablation) | GRAM-HD · FFHQ 256 | FID: 12.0 -> 11.8 | Improved |
| add dedicated background super-resolution network (ablation) | GRAM-HD · FFHQ 256 | FID: 13.3 -> 12.0 | Improved |
| add LR generator intermediate features as input (ablation) | GRAM-HD · FFHQ 256 | FID: 14.0 -> 13.3 | Improved |
| add sub-pixel convolutions (ablation) | GRAM-HD · FFHQ 256 | FID: 14.4 -> 14.1 | Improved |
| use ESRGAN base network instead of StyleGAN2 decoder (ablation) | super-resolution CNN · FFHQ 256 | FID: 18.2 -> 14.4 | Improved |
Later papers that compared with this method
1 later accepted paper reports GRAM-HD as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 2024 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 4 on most shared results.
- StyleNeRF — StyleNeRF: A Style-based 3D Aware Generator for High-resolution Image Synthesis (ICLR 2022) · better on 2 of 4 shared results
- GRAM — GRAM: Generative Radiance Manifolds for 3D-Aware Image Generation (CVPR 2022) · better on 3 of 3 shared results
- StyleSDF — StyleSDF: High-Resolution 3D-Consistent Image and Geometry Generation (CVPR 2022) · better on 2 of 3 shared results
- EpiGRAF — EpiGRAF: Rethinking training of 3D GANs (NeurIPS 2022) · better on 3 of 3 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
- GRAM: Generative Radiance Manifolds for 3D-Aware Image Generation (CVPR 2022 · similarity 0.92)
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 2024 · similarity 0.91)
- Learning Detailed Radiance Manifolds for High-Fidelity and 3D-Consistent Portrait Synthesis From Monocular Image (CVPR 2023 · similarity 0.90)
- 3D-Aware Image Synthesis via Learning Structural and Textural Representations (CVPR 2022 · similarity 0.90)
- G3DR: Generative 3D Reconstruction in ImageNet (CVPR 2024 · similarity 0.87)
- Dual Mapping of 2D StyleGAN for 3D-Aware Image Generation and Manipulation (Student Abstract) (AAAI 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.
