Slashdot Mirror
Google Experiments with Video Blogging
PunkOfLinux writes "TechWeb
has an article about Google's plans to start a video service that sounds similar to Picasa. Excerpt: 'While there's no formal announcement yet, Google co-founder Larry Page said Monday that the well-known search engine concern would soon let the general public upload self-produced videos to Google's servers, partly in an effort to learn more about how to more efficiently search and display information about video-based data.'"
For the majority of the pr0n sites out there? I know Googling is much easier (and cheaper).
No, but it is really hard to search for a specific clip in your 250 tape collection, much less the world's collection.
Kinda curious to see what the brains at Google will be able to do with this.
See my journal for slashdot ID's by year. Mine created in 2005. http://slashdot.org/journal/289875/slashdot-ids-by-year
I'd really prefer they host and search audio. Would be so much more useful.
Do you think that when you upload the video clips that you'll have to input some information about the video in order to allow it to be searched or will google tap into the vast resources of it's server farm and try to run a speech to text app and record data themselves?
I don't really know that text to speech would be a feasible option to catalog the audio contents of a file, but it would be interesting if they could implement some type of automatic content cataloging system. I suppose that if this is just going to be for video blogging that it's really not as interesting as I had first thougt, but google does always seem to try and advance what is possible.
Even just being able to post video for the world to see presents us with an interesting opportunity, but I'd love if there were something more behind this.
Those who know, do not speak. Those who speak, do not know. ~Lao Tzu
I hope they try to make the system more robust before adding resource hogging features. Note the message from status.blogger.com just today:
Monday, April 04, 2005
Some users may be experiencing an unexpected Blogger outage right now; we're looking into it and will post updated info soon. Thanks for your patience.
Posted by Eric at 10:26
If the thing is routinely failing with plain text, which it is, how is it going to work with video? - rhetorical question.
Woot community access TV for the internet!
Are they going to be using image recognition or just a boring metadata search of the video? If it's metadata , what would be stopping me from saying this video of a woman blowing a horse is just an educational video on animals or an episode of the simpsons? I wonder what kind of powerful algorithms are brewing behind their doors to tell the difference between a penis and a hot dog
*roll eyes*
Where are these innovations of which you speak? Google sells ads. They are not some second-coming of anything, including Xerox. They sell ads. What exactly "new" have they come up? I'll give you PageRank...but other than that? Why the hype?
This reminds me a little bit of a rather neat system I came across the other day, Video Google (despite the name, I don't think it has anything to do with the Google company). It doesn't use metadata or cheats like that, but rather uses image analysis to identify recurring objects and scenes.
They have a demo on their web site where you can select a portion of a video frame, and it'll show you all the places in the movie where the algorithm thinks that snippet shows up. Some other cool examples are displaying the appearances of a clock from 'Groundhog Day," and a recurring poster from 'Run Lola Run.' A research paper with more details is available here.
The abstract:
We describe an approach to object and scene retrieval which searches for and localizes all the occurrences of a user outlined object in a video. The object is represented by a set of viewpoint invariant region descriptors so that recognition can proceed successfully despite changes in view-point, illumination and partial occlusion. The temporal continuity of the video within a shot is used to track the regions in order to reject unstable regions and reduce the effects of noise in the descriptors.
The analogy with text retrieval is in the implementation where matches on descriptors are pre-computed (using vector quantization), and inverted file systems and document rankings are used. The result is that retrieval is immediate, returning a ranked list of key frames/shots in the manner of Google.
The method is illustrated for matching on two full length feature films.