Training a Neural Network for Target device in TensorFlow

This project aims to leverage the power of TinyML by combining TensorFlow and STM32CubeMX to deploy a neural network model on the STM32 Discovery Board. The main goal is to train a neural network using TensorFlow and Keras, fine-tune its performance through optimization techniques, and then convert it into a TensorFlow Lite model. The TensorFlow Lite model is seamlessly integrated into the STM32CubeMX development environment using the XCube.AI package. The IAR Embedded Workbench serves as the tool for compiling and building the generated code. The project involves multiple stages, including the generation of a suitable dataset, training the neural network model, evaluating its performance, and finally deploying it on the STM32 Discovery Board. Throughout the project, valuable insights are gained into essential machine learning concepts such as data pre- processing, model training, evaluation metrics, and the deployment of models in resource-constrained environments. This practical experience fosters a deeper understanding of the end-to-end process of TinyML and equips the developer with valuable skills in machine learning, deep learning frameworks (TensorFlow and Keras), efficient dataset handling, thorough model evaluation, and effective deployment strategies. By overcoming challenges encountered during the project, the developer gains problem-solving abilities and becomes proficient in adapting machine learning models for real-world applications, especially in the context of resource-constrained devices. Ultimately, this project significantly enhances the developer's knowledge and skills in the field of machine learning and provides a solid foundation for future TinyML endeavours.