Overview / Usage

Emotion recognition using convolution neural networks using fer2013 data-set which can be found in kaggle. Using 3 layered cnn model to predict 7 different emotions by a picture of 48 * 48 pixel size.

Methodology / Approach

Here, I have used three layered convolution neural network using Keras layer library and TensorFlow at the back end.The input is converted into 48*48 pixel size image and gray scaled. Now the input image is passed to CNN model. CNN model is trained by using facial emotion recognition data set produced in 2013 which is available in Kaggle. This dataset contains around 31 thousand images i.e. an image information is stored in array which contains the gray scale intensity for each 2304 pixels of a single image.

Technologies Used

Jupyter Notebook
Deep learning Libraries:
Keras
Tensorflow

Repository

https://github.com/adizz2407/EMOTICON