Posted by Michael Tyka, Artists + Machine Intelligence

Discovering and getting started with Machine Learning can be daunting. Perhaps you have a vague project idea and are looking for a place to start and adapt from. Or you’re looking for inspiration and want to get a sense of what’s possible.

Today we’re launching Seedbank, a place to discover interactive machine learning examples which you can run from your browser, no set-up required. Each example is a little seed to inspire you that you can edit, extend, and grow into your own projects and ideas, from data analysis problems to art projects.
Recently Google has been releasing many examples of Machine Learning code in the form of Colab notebooks. Colaboratory is Google’s hosted Jupyter notebook environment. Colab allows you to run the code directly through your browser using a free GPU provided by Google, with no setup required.

Examples include the new getting started experience on tensorflow.org, Machine Learning Crash Course, research articles on distill.pub as well as an increasing number of tutorials on tensorflow.org such as machine translation. TF Hub also provides a variety of pretrained machine learning modules ready for application which are often accompanied by Colab notebooks, exemplifying their use and making it easy to get going.
Seedbank now provides a place to search for Colab-powered Machine Learning examples. You can use the top-level categories to narrow your exploration and search for keywords inside of notebooks. Each seed has a preview that lets you quickly assess if you want to explore further. Once you click through to the Colab notebook, you’ll be immediately connected to a GPU kernel and can start working through the example or tutorial. For now we are only tracking notebooks published by Google, though we may index user-created content in the future. We will do our best to update Seedbank regularly, though also be sure to check TensorFlow.org for new content.
But the best part is that Colab lets you edit the notebooks, save copies to Google Drive, and share those derivatives with your friends or on social media — all the while you can keep using the Colab GPU for fast training and inference. You can also read data from Google Drive which makes importing large datasets a snap. Naturally we have examples of how to do that on Seedbank.

Happy exploring — hopefully you will be able to find and plant the seeds of your ideas even faster now!

Acknowledgements

Seedbank was built with contributions from Mike Tyka, Sures Kumar Thoddu Srinivasan, Chris Boudreaux, Simon Doury, Harini Krishnamurthy, Mike Dory, Gabriel Schubiner and Kyle Pedersen and with support from the Artists & Machine Intelligence and Colaboratory teams.