Overview / Usage

The emotion detection model has several use-cases and can be used for making an emotion based music player. It can also be used in conferences, seminars and classrooms to determine the emotions of the subjects and transforming it into something interesting. It can also be used widely for security purposes.

Methodology / Approach

The step-by-step workflow from preprocessing dataset to finally implementing the model on video camera feeds is given below:

1. The faces were detected in the images using OpenCV and were cropped.
2. The cropped faces were grayscaled and reshaped to 48*48 size to reduce the complexity.
3. Then a Convolutional Neural Network model was built to classify the seven different emotions
   using Intel’s Optimized Tensorflow and Intel’s Optimized Python .
4. Model was trained using Intel’s DevCloud and an accuracy of 75% was achieved on Training
   set and 74% on Test set.
5. Save the model in a .h5 file.
6. Load this model and implement the model on real-time webcam video feeds to classify
   emotions.

Technologies Used

Intel Optimized Tensorflow and Keras library and Intel’s optimized Python on
Intel AI Devcloud.

Repository

https://github.com/sbrocks/emotion