Multi-Modal Learning with Explainable Artificial Intelligence for Crop Analysis: A Comprehensive Review

Abstract

In modern agriculture, artificial intelligence (AI) is doing excellent work in crop monitoring, crop disease detection, crop yield prediction, and crop stress assessment. Various techniques such as deep learning, generative models, vision transformers, explainable AI (XAI), multimodal fusion, etc., have helped in building intelligent crop analysis. This paper provides the comparative crop analysis of various crop species, data modes, and environmental conditions for the review of benchmark studies for the current framework, experimental methodologies, and datasets. The important challenges and open issues identified are limited field datasets, class imbalance, dataset bias, high computational complexity, privacy concerns, etc. Based on these, we suggested future work that can include foundation models, digital twin techniques, federated learning, multimodal frameworks, and interpretability architecture. This review provides a review for creating reliable, scalable, and sustainable AI-driven crop analysis systems. In addition to that, the survey seeks to give researchers and AI practitioners a comprehensive analysis of the current situation.

Plant phenotyping relevance

作物の病害・収量・ストレス評価を対象に、AI手法、データモード、ベンチマーク、実験方法、データセットを体系的にレビューしており、植物表現型取得・推定手法のレビューが中心です。

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.