AI Paper Atlas

Research map / Detection and segmentation / CNN features

Bridging the Geometry Mismatch: Frequency-Aware Anisotropic Serialization for Thin-Structure SSMs

Jin Bai, Huiyao Zhang, Qi Wen, Ningyang Li, Shengyang Li, Atta ur Rahman, Xiaolin Tian

ECCV 2026 · Detection and segmentation › CNN features · approach: Domain-specific segmentation and detection · 1 citations (OpenAlex)

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 stage-adaptive gating (EEGG) in LGB (ablation)FGOS-Net · DeepCrackmIoU: 87.81 -> 91.29Improved
use hybrid EEGG instead of all-ECA gating (ablation)FGOS-Net · DeepCrackmIoU: 88.54 -> 91.29Improved
use T=3 evolution steps instead of 1 (ablation)ASGP · DeepCrackmIoU: 88.93 -> 91.29Improved
increase evolution steps to T=7 (ablation)ASGP · DeepCrackmIoU: 91.29 -> 88.30Hurt
use 64 probes instead of 128 (ablation)ASGP · DeepCrackmIoU: 87.78 -> 91.29Improved
use 64 probes instead of 32 (ablation)ASGP · DeepCrackmIoU: 90.80 -> 91.29Improved
add Hilbert on HH path instead of Raster (ablation)FA-Scan · DeepCrackmIoU: 89.12 -> 90.46Improved
align sub-band scan assignment (LH->H, HL->V) instead of mismatched (ablation)FA-Scan · DeepCrackmIoU: 88.74 -> 90.46Improved
use aligned FA-Scan instead of Hilbert on all sub-bands (ablation)FA-Scan · DeepCrackmIoU: 89.46 -> 90.46Improved
add DWT decomposition without scan alignment (ablation)Standard Mamba encoder · DeepCrackmIoU: 85.98 -> 86.44No clear effect
replace Raster with 4-direction Cross-Scan (ablation)DWT baseline · DeepCrackmIoU: 86.44 -> 88.15Improved
add FA-Scan (with LGB) instead of Raster (ablation)DWT baseline · DeepCrackmIoU: 86.44 -> 90.46Improved

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.

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

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