Edge Impulse Deep Dive Session

Description:

Deploying advanced vision AI models on embedded systems doesn’t need to be complex or time-consuming. Join us in this hands-on session where attendees will walk through two workshop examples of deploying vision-based models—each taking under one hour using the Edge Impulse machine learning development platform. In the first part of the session, attendees will learn how to collect a high-quality dataset to train and deploy real-time object detection on low-power microcontrollers. In the second part of the session, attendees will construct a multi-stage machine learning pipeline for pose classification:

Workshop #1 – FOMO: Real-Time Object Detection on Low-Power Microcontrollers (~ 60 Minutes) 

Edge Impulse FOMO (Faster Objects, More Objects) is a novel machine learning algorithm that brings object detection to highly constrained devices. It lets you count objects, find the location of objects in an image, and track multiple objects in real-time using up to 30x less processing power and memory than MobileNet SSD or YOLOv5. In this exercise, attendees will learn how to collect a high-quality object detection dataset to train and deploy a FOMO model to a microcontroller like the Arduino Portenta H7 + Vision shield.

Workshop #2 – Pose Classification: Multi-Stage Inference in an Embedded Device (~ 60 Minutes) 

Construct a multi-stage machine learning pipeline that captures image data and use the TensorFlow Pose Estimation model to identify joint locations on a human body and classify poses using those features. To accomplish this project, we will wrap the pose estimation model in a custom Edge Impulse block so that both the pose estimation and classification models can be easily deployed to an embedded device after training.