Recently, there has been a growing focus on artificial intelligence (AI), particularly on a specific type known as machine learning. This technology is extensively utilized in smart devices. However, traditional machine learning methods require large models and substantial computing power to operate effectively. In response to this challenge, the concept of "tiny machine learning" (TinyML) emerged. TinyML aims to enhance machine learning capabilities on small and resource-constrained devices by enabling them to perform actions locally, without relying on cloud-based resources. TinyML represents a relatively new and exciting field that combines machine learning with embedded devices, which are often compact and energy-efficient. The primary goal is to ensure that even when transitioning from high-end systems to low-end devices, the performance and accuracy of machine learning are well-maintained. The focus of this paper is to provide an overview of the current state-of-the-art in TinyML, covering various aspects such as background information, software implementations, hardware requirements, and the frameworks utilized in this emerging field.
ABBAS et al. (Fri,) studied this question.