Above Deep learning model is for image classification of two categories with the accuracy of 99% on training set and 75% on the test set and if you want to increase the accuracy of test set you can just add one more layer of convolution to the CNN model and the model's accuracy will increase to 80%. The predict method will be updated soon.