Volumes of video content created and distributed is increasing exponentially. Anyone who is distributing content needs to increase their content compliance efforts to ensure the content is appropriate for the viewers. Facebook alone has an army of 7500 content moderators who are moderating the content people are uploading to the service.
Recently, there has been an increase in image recognition services that highlight the inappropriate image using AI trained to spot nudity and sexual imagery. It is impressive what the AI algorithms can do with images, but the image recognition does not solve comprehensively the challenges video-based businesses are facing with their compliance teams.
Media companies have to look for inappropriate material from the audio and visual content, and the decision process is more complex than inspecting plain images.
We have found that media industry is looking for a solution to make their content compliance teams more effective by letting AI to review the content and highlight the inappropriate pieces (e.g. nudity, violence, bad and hateful language, substance use and more).
We at Valossa have built the most advanced audiovisual content compliance AI solution in the market for video businesses. It tracks not only visual sex, nudity, violence and substance use, but also detects elements of inappropriate language used. The AI creates time-based metadata that helps compliance teams to work with video sequences instead of still frames. Deep scene-level search solution helps people to discover content using advanced audiovisual recognition engine enabling rapid deep dives into the segments to inspect its appropriateness for distribution.
We believe our solution is a game changer for the content reviewing process in the digital age.
Our customers are often asking about how do we compare against the competition. We are constantly benchmarking ourselves as we want to deliver our promise of being the market leading solution for video content intelligence.
Dataturks recently wrote a great blog post called “How AI/ML algorithms see nudity in images” comparing a few of the available image APIs for recognizing nudity. Dataturks test included images that either contained or didn’t contain nudity.
Our product focus is mainly in video recognition, but we also offer image recognition API to our customers upon request. We wanted to test how well our nudity recognition works with the Dataturks nudity images benchmark and compared the results against the competition.
After running the tests, we are excited to report that our hard work has paid off and we have the market-leading nudity recognition solution for visual imagery.
Microsoft, Google, AWS and Clarifai results in Table 1 are directly taken from Dataturks article and we added Valossa results to the same comparison.
Precision shows the percentage of relevant instances among the retrieved instances and recall shows the percentage of relevant instances that have been retrieved over the total number of relevant instances. F1 is a measure of overall accuracy showing the harmonic average of the precision and recall.
In this benchmark, Valossa delivers highest precision, recall and F1 values for the recognition of nudity. However, nudity is only a subset of our comprehensive solution for video content compliance. Valossa AI recognizes nuanced elements of violence (both acts and impacts of violence), substance use, guns, weapons, explosions, video structural elements, bad language, and compiles recognitions into reports with time-based metadata and interactive tools for inspection.
Our solution’s superiority over existing image recognition APIs is highlighted with a comprehensive audiovisual content profiling via object and sound recognition, providing exact start and end point markups for each recognized occurrence and indexing it for the powerful scene-level search engine. The solution helps content compliance teams to create logs and take actions based on the results.
Valossa has the most comprehensive and effective content compliance solution in the market. Don’t just take our word for it. We encourage you to log in to portal.valossa.com and try it out yourself.
CSO Sami Niska