Deep learning has been the major buzzword amongst the tech community. Python has been the go to language or sometimes R as well. Amongst Python, TensorFlow is the most popular deep learning framework
But things are changing these days, but how?
Introducing TensorFlow JavaScript or TF.js. Javascript spin on TensorFlow, the language of the web can now do Machine Learning! Yes you heard that right.
Source: https://towardsdatascience.com/how-to-train-a-neural-network-on-chrome-using-tensorflow-js-76dcd1725032
But why do we need to do ML in JavaScript? The upsides are many!
Since everything is now happening at the browser itself, The security is tremendous as nothing leaves the browser anymore.
What else? Since JavaScript is so lightweight and easy to understand and in close to python its a pretty good way to get started into Deep learning.
How does it work though? What are tech aspects of TF.js, well look below
Source https://blog.tensorflow.org/2018/03/introducing-tensorflowjs-machine-learning-javascript.html
But who should learn TensorFlow JS?
Its for everyone! You are web-dev? Perfect this is your place to start learning Deep learning/Machine learning. You are an ML engineer looking to deploy your ML models? You can start here, TF.js is the easiest way to deploy your models and even used by companies like LinkedIn!
You are a Python Expert and want to learn Deep learning? This is again a pretty good place to start!
If you aren’t any of these above, then that’s another great reason to learn TF.js! You could start with learning TF.js, you’d learn ML, if you don’t end up liking ML, you could continue the software engineering career path! JS is the most common and best language these days to continue on this path.
Having talked about what and why of TF.js is, let’s double click on its details.
Capabilities of TF.js
TF js has got an enormous amount of capabilities.
- Write and Run your model in the browser.
- Use your Python trained model in the browser
- Create Node JS based web-apps.
As apparent above, TF.js can be used in variety of ways. You could want to embed ML in your already built web-app. Well that’s simple just import tensorflow js in your script and add TF Js code. You could also want to use your python model that was trained using Keras or TensorFlow, that’s easy too!
Now that we have talked about TensorFlow JS and its capabilities, how about we talk about awesome things people have built using TensorFlow JS?
Well, a number of amazing things have been developed using TF.js People have made computer vision systems using TF.js, LSTM ones too, Even GANs.
Here are few fun demos to check out:
https://storage.googleapis.com/tfjs-examples/webcam-transfer-learning/dist/index.html
https://experiments.withgoogle.com/move-mirror
A number of companies already use TensorFlow JS, a few case studies can be found here
https://www.tensorflow.org/about/case-studies?filter=TensorFlow.js#tf-filters
You want to dig more and learn more about TensorFlow JS? Well I have authored a book on it its available here at
Amazon https://amzn.to/3Rkp4Qk
Flipkart https://bit.ly/3d1ds6l
Author - Umang Sharma