Skip to content

Latest commit

 

History

History
19 lines (10 loc) · 981 Bytes

File metadata and controls

19 lines (10 loc) · 981 Bytes

Tensorflow MNIST example

This web application uses a background worker in Heroku, to learn identification of handwritten single digits, using the MNIST handwritten numbers dataset. The Tensorflow code is based on the "MNIST for ML Beginners" tutorial, from Tensorflow's creators.

I've made the 'web' process return a job id instead of the actual job result. The job id can be saved by the client, and used subsequently to long-poll the web process. Once the worker is done, when queried with the job id, the web process returns the result.

Live demo

See it working here

Wait about 30 seconds, then click the link. If it's still not done, wait some more and click again.

I've also added app.json and this readme so you can:

Deploy to Heroku

By clicking the button below.

Deploy