A convolutional neural network framework for classifying inappropriate online video contents

Tanatorn Tanantong, Patcharajak Yongwattana


In the digital world, the Internet and online media especially video media are convenient and easy to access. It leads to problems of inappropriate content media consumption among children and youths. However, measures or methods to control the inappropriate content for children and young people are still a challenge for management. In this research, an automated model was developed and presented to classify the content on online video media using a deep learning technique namely convolution neural networks (CNN). For data collection and preparation, the researchers collected video clips from movies and television (TV) series from websites that distribute the clips online. It consists of different types of content: i) sexually inappropriate content; ii) violently inappropriate content; and iii) general content. The collected video clip data was then extracted into frames and then used for developing the automatically-content-classifying model with algorithm CNN, analyzing and comparing the result of CNN model performance. For enhancing the model performance, a transfer learning approach and different regularization techniques were adopted in order to find the most suitable method to create high-performance modeling to classify content in video clips, movies and TV series published online.


Convolution neural network; Deep learning; Pre-trained model; Transfer learning; Video content classification;

Full Text:


DOI: http://doi.org/10.11591/ijai.v12.i1.pp124-136


  • There are currently no refbacks.

Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

IAES International Journal of Artificial Intelligence (IJ-AI)
ISSN/e-ISSN 2089-4872/2252-8938 
This journal is published by the Institute of Advanced Engineering and Science (IAES) in collaboration with Intelektual Pustaka Media Utama (IPMU).

View IJAI Stats