Research map / 3D vision / Gaussian splatting
Instant GaussianImage: A Generalizable and Self-Adaptive Image Representation via 2D Gaussian Splatting
Zhaojie Zeng, Yuesong Wang, Tao Guan, Chao Yang, Lili Ju
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 |
|---|---|---|---|
| predict Gaussian opacity instead of directly predicting Gaussian colors (Color Field) (ablation) | Instant GaussianImage · Kodak / Div2k training curves | training convergence and stability | Improved |
| predict Gaussian colors directly instead of opacity (Color Field) (ablation) | Instant GaussianImage · Kodak / Div2k | final PSNR | No clear effect |
| dither with the learned position probability map instead of image gradients (ablation) | Instant GaussianImage · Kodak / Div2k | PSNR and training stability | Improved |
| include ellipse features in the MLP input (ablation) | Instant GaussianImage · Kodak / Div2k feature ablation | final PSNR | Improved |
| include triangle and color features in the MLP input (ablation) | Instant GaussianImage · Kodak / Div2k feature ablation | training convergence and stability | Improved |
| use network-based Gaussian initialization instead of GaussianImage random initialization | Instant GaussianImage · Kodak and Div2k, visual comparison at 2s | rendering quality (qualitative) | Improved |
| use CDF sampling with a fixed number of Gaussians when entropy is dominated by low-entropy regions | CDF sampling · simulated entropy distributions | sampling density | Hurt |
| use CDF sampling with a fixed number of Gaussians when entropy is dominated by high-entropy regions | CDF sampling · simulated entropy distributions | sampling density | Hurt |
Closest papers by meaning
- Large Images Are Gaussians: High-Quality Large Image Representation with Levels of 2D Gaussian Splatting (AAAI 2025 · similarity 0.86)
- Splatter Image: Ultra-Fast Single-View 3D Reconstruction (CVPR 2024 · similarity 0.85)
- DisC-GS: Discontinuity-aware Gaussian Splatting (NeurIPS 2024 · similarity 0.84)
- GaussianSR: High Fidelity 2D Gaussian Splatting for Arbitrary-Scale Image Super-Resolution (AAAI 2025 · similarity 0.84)
- IBGS: Image-Based Gaussian Splatting (NeurIPS 2025 · similarity 0.84)
- Superpoint Gaussian Splatting for Real-Time High-Fidelity Dynamic Scene Reconstruction (ICML 2024 · similarity 0.84)
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.
