A Deep Learning Based Application for Recognition and Preventing Sensitive Image.

Exploding children to sexual content that is available from the Internet is a persistent and uncontrollable problem for parents. The problem consequences are skewing children's innocent perspective or encouraging young people to aggression and discriminatory behavior. Therefore, our team finds it is necessary to build an application that will work as a filter of adult content images for different internet platforms. This research aims to develop a web browser extension to detect pornographic photos on a website and censor them. This work is done based on the image classification technique. We use deep learning technology to build image classifiers that can be executed in real-time. Our approach is implementing the most miniature model from different architecture families, then comparing the performance of each model to find the best model that balances accuracy and speed. Our approach is implementing the most miniature model from different architecture families, including ResNet, MobileNet, GoogleNet, and EfficientNet, then comparing the performance of each model to find the best model that balances accuracy and speed. An artificial intelligence server performs core processing of the classifier. It receives the address of the image to be classified from the client for processing and then returns the censored image.