Integrate Image Preprocessing into prediction models

[juhimanishjain]

• Integrate the image preprocessing into the existing prediction models

[juhimanishjain]

The model.py files for all three models (v001, vit, and fairface) have been updated to include a shared preprocessing pipeline. This includes grayscale conversion, image validation (removing rows with pixel size not equal to 48x48), and enhancement using sharpening, denoising, and bilateral filtering. The final image is resized to (1, 1, 224, 224) for ONNX inference.

In addition, an evaluate_models.py script has been added. It loads cleaned data from the SQL database (with image paths and true labels), runs predictions using all three models, and stores the results (including confidence scores) in a CSV file for evaluation purposes.

PR #35