Effective inappropriate images and videos filtering

Topic Description

In this digital age, children grow up with the Internet. They use the Internet for leisure, social and education purposes. While they can find a lot of useful information, there are also plenty of materials with inappropriate contents for children. Therefore, parental control software has been employed by the parents. In fact, this software is often pre-installed in the personal computers, mobile devices and home routers nowadays. It can limit the internet access time and filter inappropriate contents to certain extent. However, as this software filter contents mainly based on keywords and domain names, they are not effective in filtering images and videos. Furthermore, the way this software controlling access is very inflexible. It either completely allow or block a domain or sub-domain. Many modern web sites, such as Youtube, provide a wide variety of videos including good quality educational ones and some contains inappropriate content. This makes it very difficult for parents to decide whether to allow or block these web sites. Moreover, although it is very important to ensure children’s cyber security, many parents are not using any tools to control their children’s online access. This may be because many parents are either unaware of the importance of parental control, or unable to effectively control the web sites that children visit (UK Safer Internet, 2012). Therefore, to effectively prevent children from reaching the inappropriate contents, we need not only to educate the parents and children about the risks of the Internet, but also a smart parental control system.

Effectively filtering inappropriate contents from images and videos is extremely challenging. It involves a computational heavy video analysis for detecting certain objects or events. These objects and events can appear in a variety of environments and the videos can be captured under different lightings and against different backgrounds. These factors make the analysis difficult. Furthermore, what contents are inappropriate is subjective. Different people have different tolerance and perception. The smart filter needs to be able to learn its user’s perception of inappropriate contents.

In summary, this project proposes a development of a smart parental control system which can learn the parent’s perception of inappropriate materials and detect and block such contents in real time. The system should be very easy to use and suits parents who do not have a technical background.

Skills Required:

Image processing, deep learning and programming

Background Reading:

Report an error on this page


Patrick Wong