Deep learning in the browser!
The code used for this blog-post is here.
Last night I gave a talk at The Web Meetup on deep learning in the browser. It was mostly about how to deploy models built TensorFlow to the browser, by way of TensorFlow.js, but we did manage to try a model on the Google Colaboratory, and deploy it to the browser all in the space of about 5 minutes!
We also discussed a few of the items that I brought up in my recent post about deploying models via TensorFlow.js.
Deep learning in the browser! - Presentation at the April Web Meetup.
Here are the links to the presentation and the code:
- Presentation - Deep learning in the browser!
- Online Demo - tfjs fashion mnist classifier
- Code - https://github.com/silky/tfjs-fashion-mnist-classifier
Some of the discussion points from the audience were:
- Q: What are some real-world use-cases of TensorFlow.js? (A: I haven’t seen many applications yet, but they’re sure to exist!)
- Q: What do you think the scope is for new applications of AI in the browser? (A: Heaps! We need to play with the tooling to find out what they are)
- Q: Do the models need to be significantly smaller to run in the browser? (A: Depends if you have a GPU when you run it; but not generally, no. A bit smaller, yes.
We wrapped up with a few thoughts around the future of UX and AI. Ultimately, I think this is going to be a field of significant growth and importance, so now’s the time to get into it, if it’s of interest to you!
If you’re looking to adopt AI in your organisation, then check out our upcoming AI For Leadership course!
