Vehicle Damage Insurance Claim Verification

This project uses a CNN model to classify vehicle damages into six categories:

• Crack
• Scratch
• Tire Flat
• Dent
• Glass Shatter
• Lamp Broken

Model:

• 5-layer Conv2D CNN
• ReLU activation, Batch Normalization, Global Average Pooling
• Regularization: Dropout, Data Augmentation
• Class Imbalance handling: Class Weights

Instructions:

1. Extract data.zip (contains images and CSV) to the /data folder.
2. Run the notebook Insurance Claim Verification.ipynb.

Results:

The model achieved a strong AUC of 0.94 and precision of 0.72, demonstrating its effectiveness in differentiating between vehicle damage categories.

Notes:

• Images in the data.zip should be placed in the /data folder for the model to train correctly.