TensorFlow.js Monthly #4: A month for new highs, Google IO, and the next show & tell live stream

In this 4th TensorFlow.js Web ML newsletter, May the 4th be with you, as May appears to be the month for new milestones, more demos, and we even have our next YouTube show and tell planned with 6 great new creations from the community! Read on to learn more.

Google IO

As many of you know Google IO launches on 11th May 2022 and naturally the TensorFlow.js team shall have some goodies for you to check out at this epic online event - which is Google's biggest event of the year. So go ahead and sign in to the Google IO website and add the two talks below to your must view list if you want to get the latest scoop of what is going on in the land of Web ML. More details on each below.

TensorFlow.js: From prototype to production, what's new in 2022?

In our first talk, check our official team update for all the things that are new since last IO along with some fresh announcements around new tooling, models, and performance that I'm sure will get you all rather excited. If there's one talk you watch at IO, make sure it is this one! Sign up here to ensure you do not miss out - it's completely free to attend and will be held online this year - help us spread the love!

Web ML: Transfer learning in JavaScript with TensorFlow.js

Our second talk is aimed at intermediate level users of TensorFlow.js - maybe you have used our premade models and also started to dabble with using our API but looking to further your skills and go a little deeper? Well in this workshop you will learn how to make your very own "Teachable Machine" from a completely blank canvas. Familiarity with Tensors and basic TensorFlow.js API usage is assumed to get the most out of this workshop. So go ahead and sign up here if you feel comfortable, else if you are a beginner, take my free zero to hero EdX course to get to that level super fast. My edX course by contrast will take you through step by step everything you need to know assuming no prior knowledge in the Machine Learning space.

All time highs reached - we broke 90K weeklies!

I am also super excited to announce we recently broke a new record for our all time highs for weekly downloads for TensorFlow.js, going above 90K, as shown in the chart below (94.9K to be precise):

I think with a lot of new folk potentially joining the community over the month of May with Google IO and other events going on, dare I say we may even have a jolly good chance of breaking the 100K mark this or next month? Thank you to our wonderful community who are embracing TensorFlow.js and of course the many companies of all sizes from startups to large multinationals who use us in production. We are on track to keep our exponential growth for almost 3 years running since joining the team. Huzzah!

TensorFlow.js Show & Tell is back - 8th July 9AM PT!

Roll up, roll up, the next TensorFlow.js community Show & Tell has a date and is scheduled for Friday 8th July at 9AM PT over on the TensorFlow YouTube channel live stream. We have 6 new amazing presenters pushing the boundaries of what's possible in JavaScript and Web ML. From animatronic backpacks and retro computer games to auto encoders and super resolution, we have something for everyone. Feel free to steal the image below and share with your friends to join us live!

If you are new to the community this is something you don't want to miss as we will have 6 new people from all around the world talking about the amazing things they have managed to create using TensorFlow.js. Each session we choose the 6 best creations we have seen created since the last show, so you know the bar will be high, and there is something for everyone. So save the date for now, and come join us to get inspired and make new friends from the community (everyone is friendly) at this live event. Check this link for some of the previous interviews we have done to get a feel for the format and I hope to see some new usernames there! Don't be shy :-)

Demo time: 2 great new demos to check out

It would not be a TensorFlow.js monthly without at least a few demos, so feast your eyes on these two beauties:

Locked-Image Tuning: Adding Language Understanding to Image Models

Well a picture is worth a thousand words, and a GIF is a thousand pictures, so check the GIF below. Essentially though you have a photo on the left, and you type some text on the right, and the ML model will predict how likely the text statement is related to what is in the image on the left. Really neat research and obviously it runs in the browser with TensorFlow.js - love seeing researchers reach even more folk with Web ML.

Try it yourself live in the browser (go TensorFlow.js!) and learn more about how it was made over on the Google AI blog.

MedSeg.ai - in browser medical tooling that leverages TensorFlow.js

MedSeg.ai allows free and simple volume segmentation of organs, tissue and pathologies in?radiological images. You can segment the images manually, or let their AI models do it for you. Naturally this bad boy is powered by TensorFlow.js so your privacy is preserved as your data never leaves your machine as all the inference is performed client side in the browser via JavaScript. Below you can see their hippocampus segmentation in action for example:

Great to see even more companies from the medical / healthcare sector embrace TensorFlow.js to make use of our unique client side benefits. Go give them a follow over on Twitter if you feel like it.

Seen something cool? Send me your finest TensorFlow.js finds

There is only one of me, yep I am the only Developer Relations Engineer for TensorFlow.js (tell my manager to give me headcount), so any help pointing me to amazing TensorFlow.js finds you see on your travels is a great help. If you have made something cool or seen a demo out in the wild, be sure to tag it with?#MadeWithTFJS?on LinkedIn / Twitter / social so I can find it for a chance to be in our newsletter, future events, or even our?YouTube show and tell!

See you next month with even more great #WebML content that's #MadeWithTFJS. Cheers!

Jason