Getting Started with the Edge Impulse Python SDK

In this video, Shawn Hymel demonstrates how to use the Edge Impulse Python SDK to profile and deploy a model created with TensorFlow and Keras.

View documentation for the Edge Impulse Python SDK: https://bit.ly/ei-pysdk-yt-getting-st...

The Edge Impulse Python SDK is built on top of the Edge Impulse Python API, which allows you to interact with the Edge Impulse Studio to collect data, train machine learning models, and deploy your projects to a variety of edge and embedded hardware platforms. At launch, the SDK offers two main functions: profile and deploy. Together, these functions allow you to bring your own model (BYOM) to create edge machine learning applications.

The profile method allows you to generate an estimation of your model’s RAM, flash, and inference time requirements for a given hardware architecture. You can use this method as part of your machine learning design and training phase to determine if your model architecture will fit on a device or meet your timing requirements.

Additionally, the deploy method converts your model to one of several possible firmware options. By default, the deployment output is a C++ library that you can include in nearly any build system (assuming you have a C++ compiler). The library contains your trained model and necessary functions to perform inference. You also have the option to export your model to pre-compiled firmware for supported development boards or to a number of AI-accelerator build systems.

With the Python SDK, you can train models in the framework of your choice (such as TensorFlow, Keras, PyTorch, SageMaker, etc.) and deploy the models to your edge or IoT devices. The SDK is designed to help you construct MLOps training and experiments along with building CI/CD pipelines for automatically deploying your edge ML models.