Emotion Recognition System with Adaptive User Interface

Project Overview

This repository contains the code and resources for the "Final Project" in the ECE 5831 course.

The project focuses on to dvelop an emotion recognition system that accurately classifies facial expressions into seven categories and integrates a dynamic, adaptive user interface to enhance human-computer interaction. It includes key aspects like model development, data processing, and interactive visualization.

This project uses FER-2013 dataset that contains facial images (grayscale JPEG file with a size of 48x48 pixels.) classified into seven emotion categories: Angry, Disgust, Fear, Happy, Neutral, Sad, and Surprise.

Technologies Used:

• Python
• Tkinter - for GUI creation (file dialogs and interface components).
• PIL (Pillow) - for image processing and manipulation.
• OpenCV - for image handling and pre-processing.
• TensorFlow & Keras - for building and training deep learning models.
• NumPy - for numerical operations and array handling.
• Matplotlib & Seaborn - for data visualization (plots, confusion matrix).
• Scikit-learn - for evaluation metrics like confusion matrix and classification report

Resources

Here are the relevant links for the project:

• Presentation Slides: Slides used during the presentation summarizing the project work and key findings.
• Project Report: A detailed report describing the technical aspects of the project, the methods employed, and results obtained.
• Dataset: The FER-2013 dataset used for this project.
• Demo Video: A demo showcasing the functionality and features of the developed application.

Repository Structure

• final-project.ipynb:
  • Contains jupyter notebook code for the complete project with output.
• best_model.keras:
  • This is the saved keras model after training.
• run.py:
  • This script is used to launch the adaptive UI, upload a test image, and perform predictions. The saved CNN model is loaded within the script, and upon uploading an image, the prediction results are displayed on the UI. Test images for upload can be selected from the dataset directory available in the GitHub repository.

To execute this file, use the following command:

    python run.py

Upload the image from the dataset directory and see the prediction results.

• dataset
  • This directory contains sample test images for each class(2 images per class). This can be used while code execution(both jupyter notebook and run.py file)