Design and Behavior of Sparse Mixture-of-Experts Layers in CNN-based Semantic Segmentation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Pavlitska, Svetlana, Fan, Haixi, Ditschuneit, Konstantin, Zöllner, J. Marius
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911595379032064
author Pavlitska, Svetlana
Fan, Haixi
Ditschuneit, Konstantin
Zöllner, J. Marius
author_facet Pavlitska, Svetlana
Fan, Haixi
Ditschuneit, Konstantin
Zöllner, J. Marius
contents Sparse mixture-of-experts (MoE) layers have been shown to substantially increase model capacity without a proportional increase in computational cost and are widely used in transformer architectures, where they typically replace feed-forward network blocks. In contrast, integrating sparse MoE layers into convolutional neural networks (CNNs) remains inconsistent, with most prior work focusing on fine-grained MoEs operating at the filter or channel levels. In this work, we investigate a coarser, patch-wise formulation of sparse MoE layers for semantic segmentation, where local regions are routed to a small subset of convolutional experts. Through experiments on the Cityscapes and BDD100K datasets using encoder-decoder and backbone-based CNNs, we conduct a design analysis to assess how architectural choices affect routing dynamics and expert specialization. Our results demonstrate consistent, architecture-dependent improvements (up to +3.9 mIoU) with little computational overhead, while revealing strong design sensitivity. Our work provides empirical insights into the design and internal dynamics of sparse MoE layers in CNN-based dense prediction. Our code is available at https://github.com/KASTEL-MobilityLab/moe-layers/.
format Preprint
id arxiv_https___arxiv_org_abs_2604_13761
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Design and Behavior of Sparse Mixture-of-Experts Layers in CNN-based Semantic Segmentation
Pavlitska, Svetlana
Fan, Haixi
Ditschuneit, Konstantin
Zöllner, J. Marius
Computer Vision and Pattern Recognition
Machine Learning
Sparse mixture-of-experts (MoE) layers have been shown to substantially increase model capacity without a proportional increase in computational cost and are widely used in transformer architectures, where they typically replace feed-forward network blocks. In contrast, integrating sparse MoE layers into convolutional neural networks (CNNs) remains inconsistent, with most prior work focusing on fine-grained MoEs operating at the filter or channel levels. In this work, we investigate a coarser, patch-wise formulation of sparse MoE layers for semantic segmentation, where local regions are routed to a small subset of convolutional experts. Through experiments on the Cityscapes and BDD100K datasets using encoder-decoder and backbone-based CNNs, we conduct a design analysis to assess how architectural choices affect routing dynamics and expert specialization. Our results demonstrate consistent, architecture-dependent improvements (up to +3.9 mIoU) with little computational overhead, while revealing strong design sensitivity. Our work provides empirical insights into the design and internal dynamics of sparse MoE layers in CNN-based dense prediction. Our code is available at https://github.com/KASTEL-MobilityLab/moe-layers/.
title Design and Behavior of Sparse Mixture-of-Experts Layers in CNN-based Semantic Segmentation
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2604.13761