This repository encapsulates the culmination of a 4-week AI Capstone project using Keras, a key component of the IBM AI Engineering course. The primary objective was to develop a robust image classification model to determine if a stone is cracked or intact. The project processed a dataset of 40,000 images, with 30,000 used for training and 10,000 for validation.

Week 4 Assignment:
The final assignment involved a comparative analysis between two pre-trained models, ResNet50 and VGG16, implemented using Keras. The models were trained for 2 epochs, and the performance metrics are summarized in the table below:

Model	Training Accuracy	Validation Accuracy	Test Accuracy	Validation Loss	Training Loss
Resnet 50	91.08%	93.88%	54.19	0.0016	0.0070
VGG16	99.76%	99.81%	98.0%	7.4741e-05	0.0074

Conclusion:
Based on the performance metrics, it is evident that VGG16 outperforms ResNet50 for this specific dataset. With a validation accuracy of 99.81%, VGG16 is deemed more suitable for the task at hand. The successful completion of this final assignment resulted in a perfect grade of 100%.