AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

S2M-Net: Spectral-Spatial Mixing with Morphology-Aware Adaptive Loss for Medical Image Segmentation

Sanaullah Chowdhury, Lameya Sabrin

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Efficient image transformers

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 changedWhereMeasuredEffect
add Spectral-Selective Token Mixing (SSTM) (ablation)S2M-Net · Kvasir-SEGDice (%): 88.45 -> 96.05Improved
add Boundary-Focused Decoder (ablation)S2M-Net · Kvasir-SEGDice (%): 89.87 -> 96.05Improved
add MRF-SE local block (ablation)S2M-Net · Kvasir-SEGDice (%): 89.23 -> 96.05Improved
add boundary loss component to MASL (ablation)S2M-Net · Kvasir-SEGDice (%): 87.96 -> 96.05Improved
add core loss component to MASL (ablation)S2M-Net · GlaSDice (%): 77.32 -> 93.83Improved
add texture loss component to MASL (ablation)S2M-Net · GlaSDice (%): 88.67 -> 93.83Improved
use learned morphology-adaptive weights instead of fixed weights (ablation)MASL · Kvasir-SEGDice (%): 88.92 -> 96.05Improved
add per-sample morphology modulation (ablation)MASL · Kvasir-SEGDice (%): 89.37 -> 96.05Improved
increase spectral truncation K from 32 to 64 (ablation)SSTM · PH2Dice (%): 96.67 -> 93.11Hurt
increase spectral truncation K from 16 to 32 (ablation)SSTM · Kvasir-SEGDice (%): 89.87 -> 96.05Improved
increase spectral truncation K from 32 to 48 (ablation)SSTM · Kvasir-SEGDice (%): 96.05 -> 95.98No clear effect
train S2M-Net with MASL instead of Dice+CE (ablation)S2M-Net · EndoVis-2017 multiclassDice (%): 78.91 -> 83.43Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy