AI Image Classification Project (Academic)

Project Overview

This academic project uses Convolutional Neural Networks (CNNs) to classify facial emotions from images. The four emotion classes are: neutral, surprise, focused/engaged, and angry. The project also explores bias analysis by age (young, middle, senior).

The code has been modified to remove private or sensitive data, so you can run it with your own datasets.

---

Dataset

• Main dataset: FER-2013 (for most classes)
• Folder structure:
  • Dataset/train: training images for all classes
  • Dataset/age-bias-train: used to test for age-related bias
  • Dataset/age: additional images for bias mitigation
• Each subfolder represents one class.

---

Technologies

• Languages: Python
• Libraries: PyTorch, Skorch, NumPy, scikit-learn, Matplotlib

---

Installation

1. Clone the repository.
2. Install dependencies:

pip install -r requirements.txt

---

File Structure & Usage

Model Files

• cnnModel.py: CNN architectures (CNN, CNNV1, CNNV2)
• train.py: Train a model; outputs training logs, confusion matrix, and saves the best-fit model
• evaluate.py: Evaluate model performance and confusion matrix; supports k-fold cross-validation
• load.py: Test model on single images or custom datasets

Data Visualization

• classDistribution.py: Bar graph showing the number of images per class
• sampleimage.py: Displays random images and histograms from each class

How to Run

1. Update dataPath and modelPath in the scripts.
2. Select the CNN architecture by commenting/uncommenting in train.py.
3. Run train.py to train the model.
4. Use evaluate.py or load.py to test and analyze results.