AdapterMoE: A Two-Stage Hard-Routing Mixture-of-Experts Architecture for Multi-Crop Disease Recognition with Calibrated Rejection and Incremental Learning
提出农业多作物病害识别的硬路由MoE架构，方法新颖且实验扎实，对智慧农业有实质贡献。
【预印本,未经同行评审】针对多作物病害识别任务,提出基于确定性硬路由的两阶段 AdapterMoE 架构:RouterHead 依据最大 softmax 概率阈值分类作物并拒判非目标作物,Energy+KNN 双重门控模块捕获分布外输入;在冻结 EfficientNet-B0 主干上添加五个逐作物 Adapter,通过温度缩放校准。在 PlantVillage(5 作物 26 类)上,AdapterMoE 与最优基线 Macro-F1 差距在 0.24 分以内,但训练成本约为全网络基线的 9%,并可通过 add_crop 接口实现逐作物增量更新,避免专家坍缩,适合田间部署成本敏感的精准植保场景。
Abstract:Timely crop-disease identification is critical to food security. Multi-crop recognition suits Mixture-of-Experts (MoE), but conventional soft-routing MoE learns crop assignment freely end-to-end, letting a few experts dominate (expert collapse) with no semantic correspondence to crops, and facing high retraining costs, unstable rejection of non-target inputs, and a saturated accuracy ceiling. We shift the objective from accuracy toward a trade-off among deployment cost, scaling flexibility, and rejection stability, using deterministic hard routing. We propose AdapterMoE: a RouterHead classifies the crop and rejects non-target crops via a Maximum Softmax Probability threshold, with a dual-gate Energy+KNN out-of-distribution module catching distribution-shifted inputs; five per-crop Adapters atop a frozen EfficientNet-B0 backbone discriminate diseases, each calibrated via Temperature Scaling. Because experts are hard-isolated at the data level, the design avoids expert collapse and exposes an add_crop interface for local, per-crop updates instead of full retraining. On PlantVillage (5 crops, 26 classes), across a fair five-system comparison, AdapterMoE attains accuracy statistically indistinguishable from the best baselines (Macro-F1 within a 0.24-point band) while cutting training cost to about 9% of full-network baselines, expanding to a new crop in
Comments:17 pages, 7 figures, 14 tables
Subjects:Computer Vision and Pattern Recognition (cs.CV)
Cite as:arXiv:2608.08808 [cs.CV]
(or arXiv:2608.08808v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2608.08808
arXiv-issued DOI via DataCite
Submission history
From: PinHsun Huang [view email]
[v1] Sun, 9 Aug 2026 16:45:21 UTC (106 KB)