Research map / 3D vision / Face and shape
GaINeR: Geometry-Aware Implicit Neural Representation for Image Editing
Weronika Jakubowska, Mikołaj Zieliński, Rafał Tobiasz, Krzysztof Byrski, Maciej Zieba, Dominik Belter, Przemysław Spurek
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 |
|---|---|---|---|
| replace MiraGe with GaINeR | Gaussian-based image representation · Kodak reconstruction | PSNR: 59.52 -> 77.09 | Improved |
| replace MiraGe with GaINeR | Gaussian-based image representation · DIV2K reconstruction | PSNR: 54.54 to 62.20 | Improved |
| replace MiraGe with GaINeR under identical 500k fixed Gaussians | Gaussian-based image representation · Kodak (identical setup) | PSNR: 62.49 -> 77.09 | Improved |
| replace MiraGe with GaINeR under identical 500k fixed Gaussians | Gaussian-based image representation · DIV2K (identical setup) | PSNR: 58.45 -> 62.20 | Improved |
| use GaINeR instead of GaussianImage | Gaussian-based image representation · Kodak (identical setup) | PSNR: 30.46 -> 77.09 | Improved |
| replace MiraGe with GaINeR for MS-SSIM | Gaussian-based image representation · Kodak reconstruction | MS-SSIM: 0.9999 -> 0.9999 | No clear effect |
| replace MiraGe with GaINeR in physics-based editing | Physics-based deformation · Physics-based editing | MUSIQ: 61.03 -> 62.62 | Improved |
| replace MiraGe with GaINeR in physics-based editing | Physics-based deformation · Physics-based editing | FID: 372.41 -> 376.34 | Hurt |
| use GaINeR instead of explicit point cloud reprojection with GT depth | Single-image NVS · Hotdog NVS (GT depth) | PSNR: 8.57 -> 20.58 | Improved |
| use GaINeR instead of explicit point cloud reprojection with GT depth | Single-image NVS · Scan 114 DTU NVS (GT depth) | PSNR: 13.51 -> 18.09 | Improved |
| use Depth-Pro estimated depth instead of ground-truth depth | GaINeR single-image NVS · Hotdog NVS | PSNR: 20.58 -> 19.50 | Hurt |
| use Depth-Pro estimated depth instead of ground-truth depth | GaINeR single-image NVS · Scan 114 DTU NVS | PSNR: 18.09 -> 17.74 | Hurt |
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.
- GaussianImage — GaussianImage: 1000 FPS Image Representation and Compression by 2D Gaussian Splatting (ECCV 2024) · 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
- Unified Implicit Neural Stylization (ECCV 2022 · similarity 0.86)
- Learning Transferable Features for Implicit Neural Representations (NeurIPS 2024 · similarity 0.86)
- VI^3NR: Variance Informed Initialization for Implicit Neural Representations (CVPR 2025 · similarity 0.86)
- Implicit Representations via Operator Learning (ICML 2024 · similarity 0.85)
- Grids Often Outperform Implicit Neural Representation at Compressing Dense Signals (NeurIPS 2025 · similarity 0.85)
- MiraGe: Editable 2D Images using Gaussian Splatting (ICML 2025 · 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.
