Resolved Issue178

Description of Changes:

Augmentation Techniques for Overfitting Prevention:
- Albumentations Library: Implemented the albumentations library to introduce data augmentation during training.
- Gaussian Blur: Added GaussianNoise for introducing random blur to images. This helps the model generalize better to unseen data and reduces overfitting. Blurry images can contain slightly different information than crisp ones, forcing the model to learn more robust features.
- Contrast Adjustment: Added adjust_contrast to randomly modify image contrast during training. This exposes the model to a wider variety of lighting conditions, further aiding in overfitting prevention. By seeing images with varying brightness levels, the model learns to focus on the essential features rather than just the intensity.
Loss Convergence Graph:
- Incorporated a visualization of the training and validation loss over epochs using Matplotlib. This graph helps you understand how the model's performance is evolving during training.
- Training Progress Monitoring: By observing the loss curves, you can identify potential issues like overfitting or underfitting. If the training loss decreases significantly while the validation loss plateaus or increases, it might indicate overfitting.
- Early Stopping Consideration: You can potentially leverage this graph to implement early stopping if the validation loss doesn't improve after a certain number of epochs, preventing further training that might lead to overfitting.

Overall, these changes aim to enhance the model's ability to generalize to unseen data and potentially reduce overfitting by introducing data augmentation and providing a tool for monitoring training progress.

Additional Notes:

We can further customize the parameters of GaussianNoise (variance) and adjust_contrast (factor) to explore different augmentation strategies.

Recode-Hive / machine-learning-repos

Added new file with enhanced model for Facial Expression Recognition #315

Resolved Issue178