AI Facial Emotions Detector

Prerequisites

• torch (2.6.0) – Core PyTorch library for deep learning
• torchvision (0.21.0) – Provides datasets, model architectures, and image transformations useful for training
• numpy (2.2.3) – Essential for numerical computations and tensor operations
• opencv-python (4.11.0.86) – Useful for image processing, face detection, and preprocessing input images
• pillow (11.1.0) – Used for loading and processing image data

Setup

1. Download data fromhttps://www.kaggle.com/datasets/msambare/fer2013?resource=download
2. Make a data directory and copy the downloaded data in
3. Run the app through main.py

Workflow

1. Data layer finds and processes training and test data
2. Model layer trains and evaluates the data, adjust epoch number to increase model accuracy
3. 'emotion_cnn_weights.pth' file gets created to store already existing weights
4. Predict layer utilizes the model to make predictions
5. Main handles camera input and visuals.