TinyML Powered Drone in Agriculture Application

This research explores the groundbreaking synergy between Tiny Machine Learning (TinyML) and Unmanned Aerial Vehicles (UAVs) for precision agriculture, focusing on disease monitoring in tomato plants. The study integrates TinyML into UAVs using the Edge Impulse framework and ESP EYE 32. The research introduces a meticulous optimization process for key parameters that impact system characteristics such as energy consumption and accuracy. This optimization, applied to both model training and on-device processes, ensures minimal computing resources like memory size, energy, and processing time, while maintaining acceptable accuracy levels. The literature review highlights the promising intersection of UAVs, Machine Learning (ML), and agriculture. The methodology details the use of Edge Impulse TinyML algorithm for disease classification. In experimentation, the study emphasizes the superiority of single-labeled data over bounded boxes for plant health classification, investigates parameter modifications impact on training and testing accuracy, and presents optimal setup insights. The research achieves 90% training accuracy and 91.67% testing accuracy for binary class and 92.30% training accuracy and 80.00% testing accuracy for multi-class using single-labeled data. This work envisions the transformative potential of TinyML and UAVs in agriculture, particularly in disease identification in tomato plants, paving the way for intelligent systems with diverse applications.