Tensorflow JS Tutorial

In this tutorial, we will explore the core concepts of tensorflow's javascript api and implement transfer learning on the client-side.

Getting Started

These instructions will get you prepared for the upcoming tensorflow workshop.

This guide works best for installation on Mac OS/Linux

1) Software Prerequisites

Python

For python 2.7:

$ python --version

Make sure you have at least python 2.7.14

Alternatively,for python 3.6:

$ python3 --version

Make sure you have at least python 3.6.4

If you do not have the aforementioned python versions, check out the python installation page to install a more updated version of python.

---

Pip

Test pip version:

$ pip -V

This should echo the version of pip as follows:

pip 10.0.1

If pip is not recognised as a command, try reinstalling pip from here

---

Virtual Environment (Optional)

My recommendation is that you proceed with the following setup in a python virtual environment. This will keep everything neat and simple.

Install virtualenv:

$ pip install virtualenv

Test virtualenv version:

$ virtualenv --version

Create a virtual environment for this tutorial:

$ cd tfjs-tutorial
$ virtualenv tfjs-tutorial

virtualenv tfjs-tutorial will create a folder in the current directory which will contain the Python executable files, and a copy of the pip library which you can use to install other packages.

To begin using the virtual environment, it needs to be activated:

$ source tfjs-tutorial/bin/activate

All installations and modifications will be local this virtual environment.

If you are done working in the virtual environment for the moment, you can deactivate it:

$ deactivate

For subsequent parts, do not deactivate the virtual environment just yet. Check out this link for more details on python virtual environments.

---

Tensorflow.js

Install tensorflowjs:

$ pip install tensorflowjs

If successfully installed, you should see:

Successfully installed h5py-2.7.1 keras-2.1.4 numpy-1.14.1 scipy-1.1.0 tensorboard-1.7.0 tensorflow-1.7.0 tensorflow-hub-0.1.0 tensorflowjs-0.4.0

---

Node.js

Install Node.js version 8.9 or higher from here

---

Yarn

For this tutorial, we will be using Yarn for dependency management. Download Yarn stable 1.7.0 for MacOS, Windows, Debian/Ubuntu, CentOS/Fedora

---

Miscellaneous repositories

• Clone the current repository

$ git clone https://github.com/Sri-vatsa/Tensorflow-js-tutorial.git

• Clone the tfjs-converter repository (optional)

This repository is great to have for converting existing tensorflow models into models that can run on browser.

$ git clone https://github.com/tensorflow/tfjs-converter.git

---

2) Knowledge Prerequisites

I will go through some basic concepts of machine learning, Convolutional Neural Networks(CNN) and core Tensorflow concepts before we start coding.

Basics of Machine Learning and Deep Learning

So what is Machine Learning?

Even if you do not completely understand all of that it is okay. I will explain it in a bit more detail in person.

If you seem to get what machine learning is, how about get a deeper understanding of what deep learning is?

---

Basics of CNN

Convolutional Neural Networks are very similar to ordinary Neural Networks: they are made up of neurons that have learnable weights and biases. Each neuron receives some inputs, performs a dot product and optionally follows it with a non-linearity.They have a loss function (e.g. SVM/Softmax) on the last (fully-connected) layer to help achieve what we need from the neural network.

tl;dr its a massive fucntion that takes an image as input and outputs how likely it belongs to a particular pretrained class.

Check out this awesome link to a Stanford University course on CNNs.

---

Mobilenet

MobileNets are based on a streamlined architecture that uses depth-wise separable convolutions to build light weight deep neural networks. If you did not understand a single word in the previous sentence, it is perfectly fine. Just read this paper.

---

Useful Links

• Tensorflow JS github page
• Tensorflow JS Tutorials
• Tensorflow Tutorials (Python)
• Convolutional Neural Networks Stanford
• Machine Learning Basics (** Very Important **)
• Siraj Raval's Youtube Series
• Daniel Shiffman's Youtube Series
• Google Machine Learning Crash Course (Python)

Contributing

Found more interesting resources to share? Make a PR to this repo and we can all learn.

License

This project is licensed under the MIT License

Acknowledgments

• Tensorflow JS
• Python