Pornographic images classification using CNN methods on Android-based smartphone devices

Pornographic picture classification is challenging due to the high complexity and variety of pornographic photos, which can comprise skin color, gender, body form, pose, and background. As a result, a method for classifying photos with a high level of complexity to extract crucial information is required. One of them is the Convolutional Neural Network (CNN) whose learning result is a pornographic picture categorization model that consists of a constructed architecture and weight, which is information derived from learning outcomes. The learning process of CNN required a vast number of training data called the KIA dataset obtained from the AI research team PSTI University of Mataram. It consists of porn and non-porn images over 17900 images. As a classification engine, the model will be implemented on an Android-based smartphone. Thus, the application to discover, remove pornographic images in an android-based smartphone called Porn Away. According to the study’s findings, the CNN model has high accuracy (91%), precision level (92%), recall rate (92%), and a computing time of 6 milliseconds.