A Forensic Nudity Detector Based on Machine Learning
Danilo Coura Moreira, Joseana Fechine · 2018
Forensic examinations on computer devices become increasingly costly due to the large amount of files they can store. It is necessary to use algorithms that could detect pornographic images efficiently. We presented an approach that improved the work of Ap-Apid using a machine learning-based algorithm rather than using static rules. In addition, we used features extracted from a face detector. The AIIA-PID4 pornographic data set was used in the experimental phase. In the end, the proposed model reached an accuracy of 93.56%, outperforming the referenced works, which reached 79.1% and 85.05%, respectively.