CDNvideo: The analytical platform for an intellectual insert of advertizing

2019: Announcement

On February 28, 2019 CDNvideo company together with tvzavr submitted the analytical platform for an intellectual insert of advertizing to movies based on machine learning technologies.

According to the company, the developed solution improves targeting of advertizing on the viewer thanks to relevance of specific scenes of the movie and the advertized products. Advertizing of gadgets or the office equipment approaches scenes of work at office more, appropriate to insert advertizing of the car into the stage of races or a trip. Depending on category of scenes ("office", "house", "nature", "road" and so on) within the "smart" prodakt-pleysment it is possible to connect different groups of advertizing filmlets.

Before development of this analyzer the marking of scenes under an insert of advertizing filmlets was executed manually and borrowed till three o'clock time of the operator for one movie. The average time of a marking of the movie using the CDNvideo analyzer is 15 minutes, at the same time there is a possibility of the parallel analysis of a set of movies on different servers.

Control of system operation from the operator is retained, but it is directed already more to the aid in additional training of a neuronet. Accuracy increases in determination of the necessary scenes in process of accumulation by the analyzer of experience and feedback coupling from the operator.

The tvzavr online movie theater provides for this development the content on which a system will be able to learn to select tags for the prodakt-pleysment as much as possible precisely. Thus, content of online movie theater is a data set for neural network CDNvideo to fulfill technology in practice, to gather experience, necessary for its development.^[1]

We see potential in further development of products based on machine learning in our company and we observe increase in demand for them from clients. So far we see demand for the analysis of movies and other content of business subject which are seen by citizens of Russia with income above of an average and into which, respectively, it is possible to insert advertizing of goods of a premium class with the high average check. But we are sure that over time advertizing value on the basis of relevance of a context with the advertized goods and services will become indisputable for all participants of the market of video advertizing on the Internet. Yaroslav Gorodetsky, CEO of CDNvideo

In my opinion, it is very interesting experiment, both for tvzavr, and for CDNvideo. 2019 years more and more advertisers prefer to do the rollers under specific target audience. Therefore the advertizing targetirovannost, the higher it value is better. On the one hand, the machine learning technology allows to make advertizing at us on the platform, so to speak, more "smart" – directed to the specific viewer, specific emotions which that experiences at the time of viewing. For example, if the hero of the movie eats some tasty food – and the same moment joins advertizing of restaurant or food – chances that the viewer will remember a brand, much more. On the other hand, for neuronets of CDNvideo is an excellent experience which will help to fulfill all errors, will make technology more exact and fast. Alexander Pavlov, technical director of tvzavr

Notes