AIBuster - Image Classification & ML Comparison

AIBuster is an academic project that compares how different machine-learning approaches tackle the same image-classification problem. It focuses on experimentation: training multiple models, inspecting their feature representations, and understanding trade-offs between accuracy, interpretability, and compute cost.

Key Capabilities

• Image classification using multiple approaches.
• Side-by-side comparison of models and feature pipelines.
• Visualization and dimensionality reduction of image features.
• Interactive GUI for testing and evaluation.

Machine Learning & Feature Techniques

• Convolutional Neural Networks (CNN)
• Support Vector Machines (SVM)
• Quadratic Discriminant Analysis (QDA)
• Logistic Regression
• Local Binary Patterns (LBP) with Sobel filtering
• Principal Component Analysis (PCA)
• t-SNE for feature visualization

Implementation Details

• Modular Python structure for models, feature extraction, and evaluation.
• Scripts for training, validation, and result analysis.
• Pre-trained models and stored training histories for quick comparisons.
• GUI interface (main_gui.py) for simplified interaction.

Tech Stack

• Python
• NumPy, scikit-learn
• TensorFlow / Keras (CNN)
• OpenCV (image processing)
• Custom feature-extraction pipelines

Context: Academic project focused on applied machine learning and comparative analysis rather than production deployment.
Role: Machine Learning & Implementation
Made by: Aren Seferi, Montana S., Chrisopher M.