Implementation of Smart Security System in Agriculture fields Using Embedded Machine Learning

Abstract Tiny Machine Learning (TinyML), a branch of machine learning that focuses on the effectiveness of machine learning on extremely constrained edge machines, is flourishing. Deep learning techniques are being used more frequently lately in a variety of data-intensive and time-sensitive Internet of Things (IoT) apps. Because MCUs lack resources like RAM, deploying new methods like Deep Neural Networks (DNN) models on them has proven challenging. However, recent developments in the TinyML space promise to create a completely new class of peripheral apps. By eliminating the need for the cloud's omnipresent computing support, which uses power and presents risks to data security and privacy, TinyML paves the way for the development of original apps and services. Traditional machine learning needs a lot of processing capacity to predict a scenario. This computational capacity will be moved from high-end systems to low-end devices thanks to the TinyML method for machine learning on small devices. To keep the precision of the learning models, enable resource-efficient small edge devices to manage the training and deployment process, maximize computing capacity, and enhance dependability are some of the challenges presented by this change. Here in this paper, we propose a efficient method to detect animals near farmland for security purposes using TinyML and compared with many algorithms and their effectiveness.