Using Convolutional Neural Networks to Understand Video

Picture3

For a number of years now, there has been a steadily increasing emphasis in technological innovation of using patterns common in human biology and translating them into algorithms and networks – broadly defined by deep learning, a rapidly developing field of machine learning.  A particular approach in deep learning has to do with Convolutional Neural Network (or CNN) – a type of advanced neural network.

convolutional-neural-networks-cnn-65-638 (1)

Image courtesy of  http://visal.cs.cityu.edu.hk/research/hmlpe

CNN differs from regular neural networks by the flow or neurons. These large-scale networks roughly mimic neural connections in the human brain, meaning that they can adapt to differing data, learn from it and change with time. Through the use of layers in the algorithm, the network adjusts itself so that the more ‘experience’ it has – the more precise the results are.

Picture1

Image courtesy of  http://www.iro.umontreal.ca/~bengioy/talks/DL-Tutorial-NIPS2015.pdf

Deep learning, and CNN in particular, is commonly linked with innovations in the fields of voice and image recognition, as well as of object perception.

 “Understanding” a video

Webyclip offerings are empowered by its core technology solution that uses CNN to ‘understand’ video and perceive objects pictured in it. It quickly analyses video clip and recognizes thousand of concepts, translating them into products, scenes and people.

Picture2

The application of deep learning for visual marketing is rapidly transforming video commerce, delivering an acutely immersive shopping experience. Webyclip’s Spot & Shop [include link] solution is currently using CNN technology in ‘shoppable’ videos, allowing customers to purchase products they spot in a video, in real time, by clicking links whilst watching.

Picture3 

Solving the “Skip Video” problem for video ads

Picture4

In a similar sense, Webyclip uses CNN technology for a pre-roll and in-stream advertising, providing enhanced ad relevance. By automatically identifying products and concepts that are going to be shown in the video, our solution is capable of automatically matching an ad to particular moments and context of a video clip. An accurately matched content of an ad, seamlessly placed inside viewer’s context, represents a strong likelihood of striking a chord with the viewer, striking a chord with their wallet.

To learn more about Webyclip’s video recognition and visual marketing solutions, please visit our Website or contact info@webyclip.com

Comments (0)

Leave a comment

Your email address will not be published. Required fields are marked *