Firefox Plugin Annodex For Searching Audio, Video
loser in front of a computer writes "ZDNet Australia reports that 'Australia's CSIRO research organisation has developed a Firefox plugin named Annodex that allows browsing through time-continuous media such as audio and video in the same way that HTML allows browsing through text.' I've just checked Annodex out and it's very cool. The sample video from the Perl conference is way funny too." The catch is, the media to be searched has to be prepped first.
I got some Anime in ogg once. It was the Rurouni Kenshin OVA. It was such a wonderful format and I could switch between english/jap audio and subs just by right clicking a system tray icon.
I really wish the Anime community saw it as a viable format rather than using XVid and DivX for everything. OGG is beautiful.
Have you metaroderated recently?
Although I guess that might present a chicken/egg situation.
How could a computer possibly work out what media is sports or music videos or anime or tv shows or whatever.
That sounds like a doctorate in the making... I'd anticipate an 80% hit rate in genre classification (at least) within 6 months of research, just given those sorts of categories. It's just image recogition and classification, really, but with a fscking huge dataset (which is a good thing).
How is this innovative above a DVD "jump to a scene" menu? (honest question)
I watched the video, but all it seems to be is a system of sectioning audio-visual files into smaller chunks, and a browser that gives access to a "table of contents" that lets the user jump directly to a section.
Is the sectioning/table-of-content-generation process automated? It seems to be manual.
I think software is already available that can partially automate the sectioning of a video. It does this by detecting scene-transitions, and then offering up the "chunks" to the user for approval and labelling. I think such software is used in DVD authoring for generating the "Jump to a Scene" DVD menu.
This could be really useful for TV broadcasts, particularly news.
I think anybody doing closed captioning already has the descriptive content they need. (Others could use a similar process to create it.)
That info, combined with relatively easily-detectable scene transitions, would make it possible to automate the searchable video file creation to a large extent.
So the CC or equivalent would still have to be done manually but you'd have this extremely useful, huge searchable archive of video.
Not so easy for things that depend on the visual content as opposed to the spoken content, but for news it could be amazing.
Then watch as politicians and captains of industry squirm at the thought that their every word and twitch is available for searching...
This Like That - fun with words!