NeuroFuser

Resource-aware neuromodulation of multi-scale fusion attention for domain-adaptive semantic segmentation.

Read the Paper Public Library: NeuroFixer Project Home

Paper

NeuroFuser: Resource-Aware Neuromodulation of Multi-scale Fusion Attention for Domain Adaptive Segmentation
Serdar Erişen and André Borrmann
Pattern Recognition, 2026
DOI: 10.1016/j.patcog.2026.114167

Core idea

NeuroFuser treats multi-scale attention fusion as a controllable, resource-aware process rather than a fixed architectural choice. It uses NeuroFuser-style EncodingGate, EncodingModule, FusionBridge, and a neuromodulation controller to support stable CNN/ViT-compatible dense prediction.

EG
Per-location and per-channel gated feature calibration.
EM
Latent-grid feature alignment with dilation modulation.
FBr
Branch-normalized decoder/skip fusion.
Controller
Resource-aware modulation of fusion behavior.

Companion library advertisement: NeuroFixer

The public companion library is NeuroFixer, an installable PyTorch package for NeuroFuser Basic, reusable attention-fusion modules, CNN/ViT builders, and future community-contributed attention heads. 

pip install neurofixer

GitHub Repository PyPI Package

Citation

@article{erisen2026neurofuser,
  title   = {NeuroFuser: Resource-Aware Neuromodulation of Multi-scale Fusion Attention for Domain Adaptive Segmentation},
  author  = {Eri{\c{s}}en, Serdar and Borrmann, Andr{\'e}},
  journal = {Pattern Recognition},
  year    = {2026},
  doi     = {10.1016/j.patcog.2026.114167}
}