Video is the most popular media format and with all the footage available it becomes increasingly difficult to find an efficient way to search through it.
vidipedia started life as a DTI sponsored Research & Development project investigating human indexing of video archives. enigma was appointed project lead and worked with project partners; ITV, Imperial War Museum, British Film Institute, Codeworks Connect, C.C. Harvey and the Universities of Newcastle and Sunderland, to find a way to improve video content searching.
The amount of footage viewed and produced is dramatically rising; viewing figures continue to increase, more viewing platforms become available and production costs shrink. According to Nielsen, the average American now spends 5 hours watching TV/DVR footage per day, the UK is fast approaching this. If you add online video content to this, the popularity of video is undeniable. Approximately 20 hours of video footage is currently uploaded to YouTube every minute. When you take into account archive collections and out-take footage, that’s a lot of video content!
Much of this content has real value, whether it is of cultural importance, expensive to replicate or irreplaceable coverage of one-off events. Hundreds of thousands of hours of unique archive footage held on original film and tape is slowly degrading, to be lost forever if it is not digitised. The problem is that no-one can find an economic case to digitise it. We investigated whether we could unlock the value held in all of this video by making it more searchable.
Current video search systems can be quite crude. Essentially, video searching mirrors image archive search techniques; each video is associated with some human entered text data, which is then searched upon. A fundamental issue with this is that video varies in length, from short clips to entire reels, and classifying using a universal set of data doesn’t provide enough detail necessary to deem searching useful for uncovering value within lengthy or mixed content footage.
We needed to produce a system which supported indexing to a highly detailed level; down to the scene or even frame. It also needed to prove cost effective to footage owners.
We developed vidipedia as a unique suite of software which combines to deliver all the processes required to upload, annotate, manage, distribute and deliver video material in a form that is much more searchable.
Video is stored centrally; multiple archives can store footage on a centralised system. Core functionality is delivered via secure web-services and interfacing with remotely hosted web applications. Complex data can be stored about specific time segments (clips) in footage. Footage can be annotated by users to provide indexed data on; personal experience, feelings, sounds, emotions, visual – with the added bonus that in the future, this could be used to produce “living histories” for archives. Users can search on this data to find specific clips within any footage.
Keeping human annotation at the heart of the process allowed us to obtain good quality data. We implemented systems that would follow the model of Wikipedia; supporting altruistic annotation, as well as revenue sharing for commercial services; users annotate the material in return for a share of future viewing or licensing royalties.
• data tagging /aggregation to scene and frame level
• removes up-front costs of annotating video
• provides economic case for digitising archive video
• makes video FINDABLE
• makes video VALUABLE
A wide variety of different video based applications can benefit from the improved searchability that vidipedia offers; video search engines, public access portals, asset management systems, online broadcast platforms. To effectively service all application types, vidipedia is not a single "product".
We developed a system that is a whole series of flexible software components that can be configured to support specific needs of different video based applications. It's best to think of vidipedia as an "enabler." Its power is in its flexibility, therefore the opportunities to commercialise it are great and varied.
Our plans for the future are to find an application for vidipedia that would greatly benefit from the way it makes digitised video findable and therefore unlocks its value.
The model provides an extremely adaptable and non-prescriptive solution and enables vidipedia to service any application that can benefit from its video handling capabilities. We look forward to developing it with someone to suit their needs and to unlock the potential of vidipedia.
- A special thank you to Zoe Hartill for the blog pose. For more details please e-mail her at firstname.lastname@example.org, or visit their website: www.enigma-interactive.co.uk and view their portfolio.
Picture: A screenshot of 'Vidipedia'. A video is being shown at the top with links to other videos below.
• Subscribe to iNews
• E-mail us your ideas/thoughts
• Getting Young People Skilled Up
• Mixcloud: The Youtube of Radio
• Lets face the music and...?