Google processes a lot of data. Some estimates say that Google handle 1.2 trillion searches per annum. I don’t know, but I’d imagine 99.999% of those searches are based around language. However, in a world where an estimated 80% of data is unstructured, and the vast majority of that unstructured data is video data, it is clear that search has a long way to go.
The next stage is to be able to search video and, for reasons we will briefly touch on, often the techniques used to analyse video and to allow meaningful search of video is described as AI.
Why is it called AI? A few years ago we’d have called it: image searching; video searching or “software algorithms”, but since the success of Apple Siri, Amazon Echo etc., the general public psyche and lexicon has begun to consider any such technology “artificial intelligence”. This is even if it doesn’t have a “self-learning feedback loop” or some other mechanism of analysis enhancement over time. So, though it irks me, let’s join the bandwagon and call it AI video analysis and search as well.
What cannot be ignored is that this is exciting. Video archives that previously sat dormant can now be queried for all types of things:
- What footage of Princess Diana do we have?
- Have we ever shot a news item on that street before?
- Who has ever played number 7 for England?
Those image searching algorithms are nascent but are quickly being realised. On the other hand, “semantic video analysis” is more basic to date but will answer more and more questions such as:
- This part of the movie is a car chase
- Alert! “Look at this CCTV stream because a street fight has broken out”
- This movie is a “modern day remake of the magnificent seven”
- Cars on this road have gone from 100% driven to 30% self-driven in the past 5 years
Unstructured data (or unstructured information) is information that either does not have a pre-defined data model or is not organized in a pre-defined manner. Unstructured information is typically text-heavy, but may contain data such as dates, numbers, and facts as well.
So AI video analysis and search is beginning to be able to unlock the value in media archives, and this importantly allows better reuse of media assets for all types of analysis, documentary making, sports fan lookups, and countless other usages. This will only go from strength to strength, but, is that really what “AI in media archiving” is all about?