TinyML-Based Intelligent Plant Disease Detection for Low-Power Agricultural IoT Devices

Abstract

Tiny Machine Learning (TinyML) is a promising way of integrating real-time, low-power Artificial Intelligence (AI) in resource-poor agricultural contexts. The present study reviews the recent advances in smart agriculture using TinyML in plant disease and pest detection, precision irrigation, UAV based agriculture, weed classification and sensor data processing. A structured literature analysis was performed with 23 representative studies published from 2023 to 2026 including lightweight deep learning, transfer learning, knowledge distillation, attention mechanisms, edge computing, IoT integration and adaptive model optimization. Based on the reviewed studies, we can conclude that TinyML can enable precise and timely decisions in agriculture, while minimizing computational demands, communication latency, energy usage, and reliance on cloud-based platforms. Applications such as crop disease and pest recognition, weed detection, irrigation monitoring and UAV-based field analysis have good potential for real-time implementation on microcontrollers and low-power edge devices. The results additionally reveal that model compression, optimized architectures and domain adaptation are significant for sustaining reliable performance with restricted RAM and processing sources. In conclusion, while TinyML offers a scalable and energy-efficient solution for precision agriculture, there are still several obstacles to overcome, such as training the models to be generalizable, hardware constraints, variability of datasets, data security, and field testing across a large number of deployments.