"Find me a video with the same funny dog." The Google Cloud Video Intelligence service, launched in beta testing, will ask you to clarify: should you start searching by dog breed, its size, coat length or silly facial expression? For the new system, this is not a collection of pixels in a picture, but a complex and important object. Like everything else in this video.
The new service is built on top of the Tensorflow project using machine learning principles. The goal is to learn how to recognize any video content by its content in order to subsequently conduct an effective search for relevant queries. Be it small, specialized fragments or large whole films.
What was originally a solid video image, after processing, is divided into an array of individual objects with nominal and verb labels. They are assigned a weight or rank, in percentage terms, based on comparisons with similar queries. Information is taken from regular search queries, and the result of the check is used to increase the relevance of new results.
The more accurately the tags are placed, the higher the chance of finding the desired video, but Google is tactfully silent about the mechanisms for controlling this process. On the contrary, according to the leading machine learning specialist of the corporation, Fei Fei Li, this API is intended for large businesses, media holdings and service services. For those who need an effective way to manage content. For their own, purely commercial purposes, of course.
In the current format, the innovation is in no way suitable for implementation in custom products, everyday applications. Too cumbersome and "silly". However, the trouble has begun, and the video content search technology itself is likely to become a key tool for working on the Internet in the near future.