Performance Benchmarking of ML Models for Resource Constrained Devices

The Internet of Things (IoT) revolution has enabled the deployment of a vast network of interconnected devices that collect and transmit data. With the exponential growth in the number of IoT devices, the need for efficient data processing techniques has become paramount. Tiny Machine Learning (TinyML) represents a promising solution by enabling machine learning capabilities on resource-constrained devices. TinyML has its own role in the current connected world and is a promising AI alternative for developing intelligence at the edge or for the resource-constrained devices. This study underscores the importance of model compression and optimization techniques to enable the deployment of sophisticated models in resource constrained devices, paving the way for the integration of TinyML in various Internet of Things (IoT) applications. Also explores the necessity of TinyML in IoT; enabling technologies, tools and methods adapted in TinyML and focusing on a case study that utilizes deep learning to predict weather types. The work presents a detailed study on the development and optimization of a deep learning model for weather prediction, tailored for deployment on constrained edge devices. The findings reveals that while deep learning offers significant predictive accuracy, optimizing models for TinyML through TFLite and Edge Impulse is crucial for real-time, on-device inference in constrained environments.