Emotion recognition from face

Emotion recognition from faces with Deep Learning CNN network.

Setup

You will need:

• Python 3 and the following packages:

  • OpenCV 3
  • Keras (with Tensorflow backend)
  • Tensorflow
  • Numpy
  • Seaborn

• Try it out:

  1. prepare data
  2. train model
  3. test it
     1. predict_emotion.py script
     2. real time emotion recognition from webcam with real_time_emotion.py

Prepare data

We have to prepare the data for the training

• Download the fer2013.csv (source)

  • Move it to data folder and rename it to fer2013.csv if it is necessary
    • OR
  • Edit the config.ini file

• Run the prepare_data.py script

• Now you have the data for the training which you can find in the data folder

• X_train shape: (nb_samples, 48, 48)

• y_train shape: (nb_samples,)

After this you can start the training!

Training

Run train_emotion_recognizer.py

This will run the training and evaluate the trained model with the test data.

Predict data (Try it out)

You can easily try the trained model with predict_emotion.py script

Real time prediction

TODO

About

Gábor Vecsei

• Personal Blog
• LinkedIn
• Twitter
• Github
• [email protected]