The framework comprised four significant modules towards predicting the presence of WMODAs. The first step was preparation/preprocessing, including image acquisition and image preprocessing. Next, two data augmentation methods, ImageDataGenerator and DCGANs were implemented to increase the dataset size. Then, the normalization of images was implemented to reduce the bias in the training process. The second step was training, which CNN algorithms adjust the weights from learning features of input pictures. From the labeled training results, the CNN classifiers learned the patterns of target classes. The feature extraction technique used in the training phase was then implemented in the testing phase, but only from a single query image. Finally, feature vectors were passed to the trained classifier for the final prediction of the class. In the end, model explanation powered by LIME was used to explain the predicted image and increase model reliability.