M-3 - Enhancements & Restructuring - Model 3

[Thomasbehan]

This PR introduces a series of enhancements, architectural changes, and optimizations to the project. Significant efforts have been made to improve model performance, refine learning rates, and structure application components for better maintainability.

Key Changes

• 🔄 Reduced model complexity for efficiency while improving performance.
• ⚡️ Updated learning rates to mitigate initial training overshoots.
• 💡 Integrated Focal Loss function to bolster training on challenging classifications.
• 🎯 Refined metric priorities to emphasize Recall and AUC, aligning evaluation with diagnostic significance.
• 🛠 Introduced a new command scripts setup for streamlined data scraping.
• 🏗 Overhauled application architecture, organizing command-related files into a dedicated commands folder for clarity.
• 🚀 Implemented complex residual model architectures to enhance detection performance.
• 🧪 Added a Hyperparameter Tuning Module, leveraging TensorBoard for detailed performance tracking.

Structural Enhancements

• 📁 Restructured the application for simplicity, moving operational commands into a centralized commands folder.
• 🛠 Updated architecture to include a model layer, enhanced service layer, and operational commands, removing outdated 'core' and 'ai' directories.

Optimizations & Bug Fixes

• 🔧 Improved data augmentation and class balance handling for more effective training.
• 🐛 Fixed minor bugs, including a fix for the model saving format and test case updates.