To make it easy to share your embedded machine learning projects with the world, we’re introducing public Edge Impulse projects. Public projects give anyone a complete view of your machine learning project, from the raw data and labels to your signal processing configuration, and from your model architecture to memory usage on real devices. Other developers can also clone your project with a single click, making it trivial to create templates and example applications for novel sensors, new development boards or use cases.
Want to see a public project in action? See the Tutorial: Responding to your voice example project!
To publish your Edge Impulse project, log in to the Studio, go to your Dashboard, and click Make this project public.
This takes the current state of your current project, and makes it available on a public URL, ready to be shared with the world.
Other users can freely click through your project, seeing all data, labels, configuration and intermediate state; and can clone the project with a single click if they want to continue working on the project.
When you make a project public, your current dataset and all state is copied to a new project, so no-one can make any modifications to the project, and you can continue to work on your project in private until you publish a new version. To protect your data all your devices, keys and collaborators are also omitted from the public version. And if you have a reason to unpublish a version, just head to Versioning in the Studio.
Recap
Sharing machine learning models has always been problematic. There’s lots of raw data in different formats, part of the model definition is in code, and others in configuration, and you might have gigabytes of state in your model. With the new public projects in Edge Impulse this is now trivial. Not only is all state shared, but users can click through the complete project, see how you solved a certain problem, and they have access to the same visualizations as you had during development.
We are already using public projects heavily in our own internal documentation, and hope that this is a fantastic solution for people in the community, educators and silicon partners who want to share their embedded machine learning projects with the world. We can’t wait to see what you’ll build!
Jan Jongboom is the CTO and cofounder of Edge Impulse. He was tired of making screenshots of every configuration option.