Fundamental of Teachable Machine

Teachable Machine is a web-based tool that makes creating machine learning models fast, easy, and accessible to everyone.Educators, artists, students, innovators, makers of all kinds – really, anyone who has an idea they want to explore. No prerequisite machine learning knowledge required.You train a computer to recognize your images, sounds, and poses without writing any machine learning code. Then, use your model in your own projects, sites, apps, and more.

Starter Pack

Just go to teachablemachine.withgoogle.comand click Get Started. No need to make an account or log in. This playlist of tutorial videos are built into the tool to help you along the way. You can also check out Dan Shiffman’s Coding Train video on Teachable Machine. Three starter projects for recognizing fruit, head tilts, or sounds you make.You can currently train Teachable Machine with images (pulled from your webcam or image files), sounds (in one-second snippets from your mic), and poses (where the computer guesses the position of your arms, legs, etc from an image). More types of training may be coming soon :)

Getting start with Facial Expression Recognition

Facial expressions convey the emotional state of an individual to observer what is at the back of mind. What we do while we speak often says more than the actual words. There are six basic types of emotions 1.Happiness 2.Sadness 3.Fear 4.Anger 5.Suprise 6.Disgust.

How facial expression recognition works?

A convolutional neural network for classifying human emotions from dynamic facial expressions in real time. We use transfer learning on the fully connected layers of an existing convolutional neural network which was pretrained for human emotion classification. A variety of datasets, as well as our own unique image dataset, will be use to train the model. Finally, a live video stream connected to a face detector feeds images to the neural network. The network subsequently classifies an arbitrary number of faces per image simultaneously in real time and display the probability of emotion in range of 0 to 1. Facial expression recognition work in 2 parts:

1. Facial Detection:- The ability to detect the location of face in any input image or frame. The output is bounding box of the deteced faces.
   
2. Emotion Detection:- Classifying the emotion on the face as happy, angry, sad, surprise, disgust or fear.
   

Learn more about facial expression recognition in Courcera course.
Link to DEMO what we are going to build.

Lets Build the Model

Vist teachablemachine.withgoogle.com and clicking on "Get Started". You will see three option to create an image, audio, or pose project. For now, pick “Image Project”.

Click "Add a class" upto 6 classes. Rename "Class 1" to "Happiness" ,"Class 2" to "Sadness", "Class 3" to "Fear", "Class 4" to "Anger", "Class 5" to "Suprise" and "Class 6" to "Disgust"
Now is the most interesting step, we have to add image samples to each class with respective emotions. Like smiling, happy images in Happiness Class. We have two option Webcam and Upload to add images. For video guide click hear.
Click on Train Model and let other values as default.Your Machine Learnig model is ready! Preview it by enabling webcam or input a image.Now its time to Export Model. We can export model in 3 differnt ways. But for now we will use default Tensorflow.js. Select Upload my model to add our own model to cloud for further use.Click on Copy option from code snippet. Open new HTML template on repl.it and paste the code between body syntax.Click on Run button to see your Realtime Facial Expression Recognition working.Select Start Button to get started. Give access to your webcam. It wil take couple of seconds to start webcam and connect to model.

Hacking

There are different ways in which we can advance in Facial Expression Recognition. We can download Tensorflow model converted into keras .h5 model and load it in our Python Project.We get list for probablity of prediction for each class we add while creating Classes (Step: 4). Tensorflow Lite model to use in android application or IOT and Take a look at classses we have add. We can add more class to it like Neautral, winking left or right side,etc. In the begining the result will be not that much accurate, but as time pass it will show more optimal results.Think of where we can use facial expression recoginition can be implement in real life. Below are some new example to keep hacking going on.

Tensorflow Trained Model With Python: DEMO CODE
Advance Facial Expression Recogination: DEMO CODE
Beautify boring looking website with CSS & JS: DEMO CODE