Project Details

AI / Machine Learning 2025

AI Facial Emotion Recognition

This was an Artificial Intelligence course project. The goal was to build a Machine Learning model capable of detecting and classifying 7 basic emotions (happy, sad, angry, etc.) via a real-time webcam feed.

Python TensorFlow OpenCV AI/ML

Technical Description

The project addresses a Computer Vision problem. The biggest challenge was collecting and normalizing the facial dataset and training the model to achieve high accuracy under different lighting and angle conditions.

I used the OpenCV library to process the video, extract frames, and perform face detection. Then, each face was cropped, converted to grayscale, and fed into a trained Convolutional Neural Network (CNN) model to predict the emotion.

Through this project, I learned the complete workflow of an AI project: from data preprocessing, selecting model architecture, training, evaluating, and deploying a real-world application.

Project Type

Course Project (Artificial Intelligence)

Timeline

Approx. 3 Months

My Role

Main Developer

Team

Personal

Deliverable

Python Application (Desktop) Real-time

Implementation Process

Data Collection & Preprocessing

Collected a facial image dataset (e.g., FER2013). Used OpenCV to detect, crop, normalize image sizes (e.g., 48x48 pixels), and convert to grayscale.

Model Building & Training

Designed a Convolutional Neural Network (CNN) architecture. Used TensorFlow and Keras to train the model with the processed data. Tuned hyperparameters to increase accuracy.

Model Evaluating

Tested the model on the validation and test datasets. Used a confusion matrix to evaluate the accuracy for each emotion class.

Application Deployment

Loaded the trained model (.h5 file). Built a Python script using OpenCV to read the webcam feed, apply the model for prediction, and display the emotion result directly on the video.