This directory contains scripts and tools for training, evaluating and deploying human activity recognition (HAR) models using TensorFlow, Sklearn & STM32Cube.AI.
Under training you can find a step by step guide plus the necessary scripts and tools to train, quantize and evaluate your image classification models using WISDM dataset. Note that the dataset is not provided in the modelzoo and has to be downloaded by the user himself.
Under evaluate you can find a step by step guide plus the necessary scripts and tools to benchmark your model using STM32Cube.AI through our STM32Cube.AI Developer Cloud Services or from the local downloaded CLI, as well as support for quantizing and evaluating your pretrained IGN, SVC, and Custom models is also provided if dataset are provided.
Under deployment you can find a step by step guide plus the necessary scripts and tools to quantize your own pre-trained human activity recognition (HAR) models, evaluate it and then deploy it on your STM32 board using STM32Cube.AI.
You can also use a pretrained model from our human activity recognition model zoo, and deploy it for your tests. Refer here for all the available pretrained models.