Few-shot Metric Domain Adaptation: Practical Learning Strategies for an Automated Plant Disease Diagnosis

2024-12-25

AI Summary

Plant disease image recognition suffers from severe performance degradation under cross-environment settings (e.g., diverse fields, imaging devices, lighting conditions), especially when the target domain provides only a few labeled samples per class (~10 images). Method: We propose Few-shot Metric Domain Adaptation (FMDA), a lightweight, end-to-end differentiable framework that explicitly regularizes feature space alignment between source and target domains via distance-based constraints—enabling label-free domain adaptation compatible with arbitrary deep models. Contribution/Results: Evaluated on 223,000 real-world farm images across 20 fields and 3 crops, FMDA achieves F1-score gains of 11.1–29.3 points over source-only baselines and outperforms fine-tuning methods using equivalent target data by an average of 8.5 points. The method significantly enhances diagnostic robustness and practicality in few-shot cross-domain scenarios.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Adversarial Attacks & Robustness
• Application Domains: Web

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Systems and Infrastructure for Web, Mobile and WoT: Web applications in cross-disciplinary domains and verticals such as mixed reality, smart cities, and digital health
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Numerous studies have explored image-based automated systems for plant disease diagnosis, demonstrating impressive diagnostic capabilities. However, recent large-scale analyses have revealed a critical limitation: that the diagnostic capability suffers significantly when validated on images captured in environments (domains) differing from those used during training. This shortfall stems from the inherently limited dataset size and the diverse manifestation of disease symptoms, combined with substantial variations in cultivation environments and imaging conditions, such as equipment and composition. These factors lead to insufficient variety in training data, ultimately constraining the system's robustness and generalization. To address these challenges, we propose Few-shot Metric Domain Adaptation (FMDA), a flexible and effective approach for enhancing diagnostic accuracy in practical systems, even when only limited target data is available. FMDA reduces domain discrepancies by introducing a constraint to the diagnostic model that minimizes the"distance"between feature spaces of source (training) data and target data with limited samples. FMDA is computationally efficient, requiring only basic feature distance calculations and backpropagation, and can be seamlessly integrated into any machine learning (ML) pipeline. In large-scale experiments, involving 223,015 leaf images across 20 fields and 3 crop species, FMDA achieved F1 score improvements of 11.1 to 29.3 points compared to cases without target data, using only 10 images per disease from the target domain. Moreover, FMDA consistently outperformed fine-tuning methods utilizing the same data, with an average improvement of 8.5 points.

Problem

• Plant Disease Recognition
• Environmental Adaptability
• Limited Training Data

Innovation

• FMDA
• Plant Disease Recognition
• Few-shot Learning