Recogmission, the start-up using the multimedia and text data for cross-content search in its Picollator.com search engine, announced the first release of the automatic visual content filter based on the object detection and pattern recognition technologies.

The task of processing and filtering the adult content is a nightmare for search engines, social networks and photo sharing sites. Most important issue is that adult content can affect the corporate policy, brand recognition in the market, and sometimes lead to lawsuits.

According to the traditional approach, the adult content can be moderated manually or by keywords ‘adult’, ‘sex’, and other related tags. One can analyze the text content much easier in contrast to pure images. However, the user-generated content often does not contain appropriate keywords. Manual operations are too much resource and time consuming. In addition, the quality of content filtering depends on the attention and subjective evaluation by the content managers.

The text-based method to power the ‘safe search’ mode used by search engines can reduce the manual processing. However, it is not a smart criterion for filtering. Users describes the pictures at the websites using adult-related words, therefore the textual content depends on the descriptions made by the users. Indeed, people do not describe the photos accurately. According to Recogmission’s estimations, more than 30 percent of text descriptions and tags do not match the image content. In addition, the text-based approach often leads to serious mistakes that may have the negative impact. For instance, a website provides the information about the solution that helps to protect other services and users from the pornography. As a result, text-based filters can blacklist the website because its page contains the adult keywords.

Recogmission technology implemented on the basis of Picollator.com algorithms is going to change the traditional approach to the content filtering. Picollator content filter understands the images and recognizes the adult content in the pictures without text descriptions. The porn images or sites containing such content can be blocked for users automatically. Its engine can be integrated into any online or mobile service to detect the adult images, filter search results or generate alerts.

Picollator filter automates most tasks in the moderation process, reducing costs on the manual operations of the content managers, improving service recognition in the public environment and protecting the business against content-related claims. The solution is of interest for any company to improve multimedia content moderation. It also can protect users from the adult content appeared in e-mails, search results or on the website.