Technology
MetaLabs use cutting edge video and audio analysis technology to create deep, structured metadata for moving image assets.
Our analysis systems operate in three core areas:
1.Speech to Text:
Text is the currency of online search application, Video differs from one dimensional web objects such as image files or text pages as it has the added second dimension of time. By creating a speech to text transcript of the video, MetaLabs can build a keyword set associated with the timeline of the video element. This temporal tagging of content provides a clear view of what is happening when within a piece of video.
2.Scene Boundary Detection:
Combining a range of approaches from Keyframe logging, edge detection, histogram and luminance analysis a clear map of scene boundaries can be generated. This structural map of a video asset can be used to identify suitable ad insertion points for mid roll video advertising or companion banner ads and to create segment boundaries for streaming fragments of video content.
3.Keyword and Category Generation:
Using Natural Language Processing techniques, MetaLabs analyse text transcripts to generate and refine keyword sets relating to places, people, activities and other useful handles.
MetaLabs can customise the output of our analysis systems to generate a range of outputs including existing online search/RSS/multimedia metadata schemas or bespoke schemas used by our clients.
