Google's machine learning and artificial intelligence programs are about to get a lot smarter. At their Cloud Next conference, Google announced a new feature called the Cloud Video Intelligence API. It uses deep learning models to detect objects and themes in video content. The idea is to be able to translate a YouTube or video stored in Google's cloud into keyword descriptions about its content.
With this new feature, you'll be able to search for videos not based on user generated tags, but by the actual content they contain. For example, you could search for "dogs", "humans running", "birds flying", etc., and the API will return content that contains those entities. A demo they showed at the conference was able to identify the breed of dog in a video, and also detect that it was part of a commercial.
The tool can also determine exactly when the entity appears in the video and context of what it's doing.
The Cloud Video Intelligence API is aimed at large media corporations to better categorize and manage their content. Searching for the right video or even the right frame in that video is a time consuming process that Google hopes this new machine-learning approach can speed up.
The API is part of the larger Google Cloud Machine Learning platform. It currently has other features like vision, speech recognition, natural language processing, and others. Use cases continue to expand into new fields like content-based product recommendations and advertisements, medical-image analysis, and security monitoring.