Validation accuracy not changing while loss is decreasing in keras image classification?

Question

I have two classes of images.

• Fake
• Real

Dataset splittion detail is below.

• Total Training FAKE Images 3457

• T