Multi-Crop Disease Detection in Computer Vision for Resource-Constrained Farms: A Review

Crop diseases remain a major threat to global agricultural productivity, particularly in resource-constrained regions, where early intervention is critical to ensuring food security. Smallholder farmers often practice intercropping, creating complex field conditions that challenge conventional disease detection systems. Although computer vision and machine learning offer powerful tools for early diagnosis, most existing approaches are tailored to single-crop scenarios and rely on cloud-based or high-resource environments, limiting their applicability in the field. This review synthesizes and explores recent advances in AI-based crop disease detection with emphasis on multi-crop and intercropped systems. We provide an overview of the detection pipeline from data acquisition and preprocessing to model training and deployment. We highlight technical challenges such as dataset scarcity, lack of modality diversity, poor model generalization, and limited real-world validation. Unlike prior reviews that isolate crops and modalities, this review uniquely integrates multi-crop, multimodal and edge deployable perspectives. We identify key research gaps and outline emerging opportunities for scalable solutions, including lightweight, multimodal models, offline-capable inference, edge deployment, and the integration of conversational AI for field-level support. By centering on the realities of low-connectivity, heterogeneous farming environments, this review aims to offer guidance toward inclusive and sustainable AI-driven agriculture.