Research map / Architectures and efficiency / Attention and transformers
MAXIM: Multi-Axis MLP for Image Processing
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 inter-stage cross-gating blocks (ablation) | MAXIM-2S · GoPro | PSNR: +0.25 dB | Helped |
| Add supervised attention module (ablation) | MAXIM-2S · GoPro | PSNR: +0.63 dB | Helped |
| Add cross-stage cross-gating blocks (ablation) | MAXIM-2S · GoPro | PSNR: +0.36 dB | Helped |
| Add multi-scale supervision (ablation) | MAXIM-2S · GoPro | PSNR: +0.26 dB | Helped |
| Add local MAB branch to baseline UNet (ablation) | MAXIM-2S · GoPro | PSNR: 30.48 -> 31.52 | Helped |
| Add global MAB branch after local MAB (ablation) | MAXIM-2S · GoPro | PSNR: 31.52 -> 31.68 | Helped |
| Add local CGB branch after both MAB branches (ablation) | MAXIM-2S · GoPro | PSNR: 31.68 -> 31.84 | Helped |
| Add global CGB branch after local CGB (ablation) | MAXIM-2S · GoPro | PSNR: 31.84 -> 31.91 | Helped |
| Increase width from 32 to 64 channels (ablation) | Single-stage MAXIM · GoPro | PSNR: 31.08 -> 32.09 | Helped |
| Increase width from 32 to 96 channels (ablation) | Single-stage MAXIM · GoPro | PSNR: 31.08 -> 32.31 | Helped |
| Increase depth from 3 to 4 downscaling steps (ablation) | Single-stage MAXIM · GoPro | PSNR: 31.08 -> 31.17 | Helped |
| Increase depth from 3 to 5 downscaling steps (ablation) | Single-stage MAXIM · GoPro | PSNR: 31.08 -> 31.43 | Helped |
Closest papers by meaning
- Linear Differential Vision Transformer: Learning Visual Contrasts via Pairwise Differentials (NeurIPS 2025 · similarity 0.87)
- Bridging the Divide: Reconsidering Softmax and Linear Attention (NeurIPS 2024 · similarity 0.86)
- AsCAN: Asymmetric Convolution-Attention Networks for Efficient Recognition and Generation (NeurIPS 2024 · similarity 0.85)
- Multiscale Low-Frequency Memory Network for Improved Feature Extraction in Convolutional Neural Networks (AAAI 2024 · similarity 0.85)
- Demystify Mamba in Vision: A Linear Attention Perspective (NeurIPS 2024 · similarity 0.85)
- Breaking the Low-Rank Dilemma of Linear Attention (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.