Convolutional Neural Network (CNN) has proven to be the state-of-the-art method to extract a great deal of features from images. Based on convolution operation in multiple dimension space, convolution layer extracts feature from an image according to the filter kernel, creating stacks of feature maps that will be passed to another classifier model. CNNs on the shallow depth detects simple features and sequentially detects more complex features as deeper the network goes. The implementation of CNNs has various backbone architectures that differ in performance, depth, and number of parameters. We use keras preprocessing layers embedded to the model architecture. The preprocessing we use in this project are normalization layer (to convert pixel value from 0 – 255 to 0 – 1) and data augmentation layer.