How to count the amount of images correctly classified in keras

Question

I have tested a model using keras and i\'ve noticed a little something i dont really understand, the accuracy of my model is 0.9020000100135803 and i have 6000