Using Image Data Augmentation to Train a Better ML Model

Run Time: 18 Minutes

In this IoT Central MicroSession with Edge Impulse, learn the process of performing data augmentation on an image dataset, which includes flipping, translating, zooming, rotating, and adding noise. This process generates modified copies of the original data. See how to train a convolutional neural network (CNN) on Edge Impulse with the augmented dataset and demonstrate how it is more accurate than a model trained on the original data.

Instructor: Shawn Hymel, Senior Developer Relations Engineer, Edge Impulse

Required Hardware: None

Preparation:
Free sign-up at https://studio.edgeimpulse.com/signup
Code and dataset: https://github.com/ShawnHymel/computer-vision-with-embedded-machine-learning

Additional Resources:
A non-augmented version of the project used in the video: studio.edgeimpulse.com/public/36514/latest
An example of the project with augmented data: studio.edgeimpulse.com/public/36800/latest
Previous

Webinar Series: Fast and Fearless - The Future of IoT Software Development - Part 3 of 4: Making the Most of IoT Connectivity

Next

Using Machine Learning for Sensor Fusion