Research map / 3D vision / Depth and stereo
Color Matching Using Hypernetwork-Based Kolmogorov-Arnold Networks
Artem Nikonorov, Georgy Perevozchikov, Andrei Korepanov, Nancy Mehta, Mahmoud Afifi, Egor Ershov, Radu Timofte
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 3-layer MLP with 1D KAN (ablation) | cmKAN baseline · sRGB-to-sRGB own dataset (supervised) | PSNR: 22.61 -> 24.90 | Improved |
| output 2D spatial parameter maps instead of pooled 1D vector (ablation) | cmKAN with KAN · sRGB-to-sRGB own dataset (supervised) | PSNR: 24.90 -> 25.23 | Improved |
| add Illumination Estimator (ablation) | cmKAN C2 · sRGB-to-sRGB own dataset (supervised) | PSNR: 25.23 -> 25.51 | Improved |
| add Color Transformer with MSA (ablation) | cmKAN C3 · sRGB-to-sRGB own dataset (supervised) | PSNR: 25.51 -> 25.59 | Improved |
| add Color Feature Modulator (ablation) | cmKAN C4 · sRGB-to-sRGB own dataset (supervised) | PSNR: 25.59 -> 25.94 | Improved |
| add spatial compression on attention vectors (ablation) | MCA block · sRGB-to-sRGB own dataset (supervised) | GFLOPs: 7.99 -> 2.94 | Improved |
| add spatial compression on attention vectors (ablation) | MCA block · sRGB-to-sRGB own dataset (supervised) | PSNR: 25.92 -> 25.94 | Improved |
| generate KAN parameters with hypernetwork instead of directly predicting RGB (ablation) | cmKAN generator · sRGB-to-sRGB own dataset (supervised) | PSNR: 23.81 -> 25.94 | Improved |
| hypernetwork-generated KAN parameters instead of fixed learnable KAN layer (ablation) | cmKAN generator · sRGB-to-sRGB own dataset (supervised) | PSNR: 24.67 -> 25.94 | Improved |
| add KAN in U-Net bottleneck instead of hypernetwork (ablation) | U-Net · sRGB-to-sRGB own dataset (supervised) | PSNR: 24.33 -> 24.89 | Improved |
| use KAN instead of MLP hypernetwork target (ablation) | Generator+MLP · sRGB-to-sRGB own dataset (supervised) | PSNR: 24.08 -> 25.94 | Improved |
| use KAN instead of FFN (ablation) | Generator+FFN · sRGB-to-sRGB own dataset (supervised) | PSNR: 24.22 -> 25.94 | Improved |
Closest papers by meaning
- RGB-Multispectral Matching: Dataset, Learning Methodology, Evaluation (CVPR 2022 · similarity 0.89)
- Semantic-Sparse Colorization Network for Deep Exemplar-Based Colorization (ECCV 2022 · similarity 0.88)
- Learning a Deep Convolutional Network for Colorization in Monochrome-Color Dual-Lens System (AAAI 2019 · similarity 0.88)
- Colorful Image Colorization (ECCV 2016 · similarity 0.87)
- Learning to Structure an Image With Few Colors (CVPR 2020 · similarity 0.87)
- MangaNinja: Line Art Colorization with Precise Reference Following (CVPR 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.
