Bringing Precision to the Margins: Lightweight Machine Learning Models for Resource-Constrained Agriculture

Precision agriculture is an approach to farming through machine learning (ML) to enhance efficiency and sustainability. Resource-constrained small-scale farmers are at a disadvantage in terms of internet, computational resources, and financial resources. Traditional ML solutions require centralized systems and high-performance hardware, which makes it difficult for marginalized farming communities to reach such systems. This paper presents lightweight ML solutions customized for these environments, particularly when applied to edge-computing and localized data. It analyses existing approaches, identifies the gaps in scalability, energy efficiency, and adaptability, and proposes a new model of lightweight, decentralized ML for applications such as pest detection, yield prediction, and soil health monitoring. The vision is to democratize access to precision agriculture, enabling sustainable farming practices and resource-limited farmers around the world.