CODAM is a collaborative project funded by Innovate UK in the area of video fingerprinting and visual search. Our partners in the project are the University of Surrey and Visual Atoms. The project aims to deliver a toolkit to track and identify footage in very large video archives. This could be of significant help to broadcasters who quickly need to find a piece of footage for legal reasons, or for programme makers who want to find shots of a particular object or scene.
The two search technologies we’re working with on this project allow different types of retrieval. Video fingerprinting allows you to quickly find identical or near-identical footage in an archive of hundreds of thousands of hours of content. This is useful if you need to know whether a particular video clip has ever been broadcast. One of the strengths of this technology is its robustness to modification. Even if a piece of footage has scrolling text overlaid on it or its framerate/colour-balance has been changed it can still be accurately identified by the system.
The second technology we’re exploring is visual search. This allows you to query a database for ‘similar’ content. For example, if you want to find footage of the Eiffel Tower you could upload a video clip of the Tower and the system will return footage that closely matches it, even if the new footage is shot from a different perspective.
The project aims to overcome the three main challenges around this sort of content-matching technology. Firstly, there is the challenge of accuracy - you need to make sure the system delivers high quality, meaningful results. Secondly, there is the challenge of scale - getting systems like this to perform over hundreds of thousands (and ultimately millions) of hours of footage is a non-trivial task. Thirdly, there is the issue of user experience - you need to design a system that is fast, adaptable and easy-to-understand.
The project will complete in September 2016 with the aim of producing research results and a search product that the BBC can use internally. Our partners aim to produce a suite of search tools for commercial usage by anyone with a large video archive they want. If you have any questions please feel free to contact us at email@example.com
Tweet This - Share on Facebook
BBC R&D - AI Production
BBC R&D - Data Science Research Partnership
BBC R&D - The Business of Bots
BBC R&D - Editorial Algorithms
BBC R&D - Using Algorithms to Understand Content
BBC R&D - Artificial Intelligence In Broadcasting
BBC R&D - COMMA
BBC R&D - Content Analysis Toolkit
BBC News Labs - Bots
BBC Academy - What does AI mean for the BBC?
BBC Academy - AI at the BBC: Hi-tech hopes and sci-fi fears
BBC R&D - Natural Language Processing
BBC iWonder - 15 Key Moments in the Story of Artificial Intelligence
Wikipedia - Artificial Intelligence
BBC News - Artificial Intelligence
BBC News - Intelligent Machines
This project is part of the Content Analysis Toolkit work stream