homepage Welcome to WebmasterWorld Guest from 54.163.72.86
register, free tools, login, search, pro membership, help, library, announcements, recent posts, open posts,
Become a Pro Member
Home / Forums Index / Google / Google SEO News and Discussion
Forum Library, Charter, Moderators: Robert Charlton & aakk9999 & brotherhood of lan & goodroi

Google SEO News and Discussion Forum

This 35 message thread spans 2 pages: 35 ( [1] 2 > >     
Google unveils YouTube antipiracy tool
YouTube can match video duplicates now!
sun818




msg:3478307
 10:58 pm on Oct 15, 2007 (gmt 0)

Google unveils YouTube antipiracy tool
[news.com...]

The automated YouTube video ID system looks at all video as it is uploaded and tries to match it with a database of visual abstractions of the copyrighted material that has been provided by content owners. If the system finds a match it will either block it, post it, or--depending upon the policy specified by the content owner--put ads on it, with the revenue being shared with the content owner.

 

Robert Charlton




msg:3478326
 11:19 pm on Oct 15, 2007 (gmt 0)

Some discussion on WebmasterWorld from back in April about what form this "filter" might take...

Google: YouTube Copyright Filter Almost Ready
[webmasterworld.com...]

"Almost" in this case meant just about six months, to the day. It's been about a year since Google acquired YouTube.

weeks




msg:3478379
 1:36 am on Oct 16, 2007 (gmt 0)

If the system finds a match it will either block it, post it, or--depending upon the policy specified by the content owner--put ads on it, with the revenue being shared with the content owner.

Hmmmm, why don't they do that with all web sites they list on their search engine results?

gibbergibber




msg:3478405
 2:51 am on Oct 16, 2007 (gmt 0)

I just don't believe this will work.

Are they really going to scan each video in many different places, then compare those scans to all the copyrighted videos in existence, and repeat this process for every upload they receive? No.

The key words seem to be that it's based entirely on "a database of visual abstractions of the copyrighted material that has been provided by content owners". Are major broadcasters or film studios really going to trawl through their entire archives to build up such databases? Again, no.

At best this might stop a few high profile videos from being illegally uploaded, but they have no chance of catching more than a tiny fraction of copyright violations.

Quadrille




msg:3478442
 3:37 am on Oct 16, 2007 (gmt 0)

But it may serve its purpose - convincing copyright owners that they serious about copyright.

Ultimately, they'll continue to depend on volunteer reports - as ebay and eveyone else does too. But if this avoids them having to display 700x300 banners on each page, they'll be happy!

menial




msg:3478460
 4:08 am on Oct 16, 2007 (gmt 0)

This is what Google had to do to comply with the law, but I don't think their management is happy about it. I'm sure they could have implemented this duplicate checking feature some time ago, but they were waiting as long as they could for a reason. They realize Youtube is on its way of becoming the next Napster - once the best and most popular free file-sharing service, now a relict of the past. Who uses Napster nowadays?

People are moving to other services like metacafe or dailymotion (or dozens of others) because they know they won't be able to find anything useful (meaning - the content that violates the copyrights) on Youtube. The smaller services are more likely to get away with copyright violations than Youtube; thus, they are becoming more useful and appealing to the current and prospective users.

thecoalman




msg:3478461
 4:10 am on Oct 16, 2007 (gmt 0)

I just don't believe this will work.

Neither do I, I mentioned this in another thread. There is too many things you can do to alter the content, crop, blur... even the re-encoding process will produce different variants of the same video depending on the settings.

At best they will be able to filter some at the start until someone figures out how to get around it which of course will require them to adjust the filters.. then someone will figure out how to circumvent that... round and round she goes.

Quadrille




msg:3478474
 4:35 am on Oct 16, 2007 (gmt 0)

People are moving to other services like metacafe or dailymotion (or dozens of others) because they know they won't be able to find anything useful (meaning - the content that violates the copyrights) on Youtube. The smaller services are more likely to get away with copyright violations than Youtube; thus, they are becoming more useful and appealing to the current and prospective users.

I suspect it's an error to characterize YouTube as simply a home for aspirant copyright thieves; I'll bet 90% of their visitors are just as happy with legal content - and even if this thingy only half works, many more sources will 'sign up' to YouTube - after all, they don't want the theft sites to succeed either, do they?

I doubt that 'People' are deserting YouTube, though it may be that the bootleggers are! And the YouTube wannabe's who fail to emmulate Google's new thingy will be first in line for the lawyers, as of now.

[edited by: Quadrille at 4:36 am (utc) on Oct. 16, 2007]

menial




msg:3478478
 4:48 am on Oct 16, 2007 (gmt 0)

I'm also happy with "legal content" - as long as by "legal content" you mean illegal content that has been uploaded and present for several hours or days until it's been removed :). The Youtube competition is likely to keep their "illegal content" for more than a few hours or days and that's their best competitive advantage.

Quadrille




msg:3478491
 5:18 am on Oct 16, 2007 (gmt 0)

For now ;)

loudspeaker




msg:3478512
 5:32 am on Oct 16, 2007 (gmt 0)

If the system finds a match it will either block it, post it, or--depending upon the policy specified by the content owner--put ads on it, with the revenue being shared with the content owner.

Hmmmm, why don't they do that with all web sites they list on their search engine results?

I was thinking the same thing - why don't they allow web site owners to upload a (legally binding) form saying, effectively "this content is mine and no one else's". Then calculate MD5 sums lets say for each paragraph longer than one sentence and put them on record.

When scrapers arrive and start copying, Google can simply check the database of MD5's and will know right away this is a page stuffed with stolen (or at least "quoted") content.

If anything, implementing it for web sites is easier than for videos.

zett




msg:3478579
 7:52 am on Oct 16, 2007 (gmt 0)

Ah - the PR geniuses at Google have "unveiled" their "new" copyright protection system just in time to brush up the quarterly earnings report, and probably to soften the beatings they might receive from analysts should things turn out to be not too good...

In fact, there is a chance that this system actually might work. Music recognition services work pretty similar. You just break down a song into a big number of samples, and upon request you just look for a similar pattern of a small sample. However, with video adding the "picture" component, the amount of data that needs to be analyzed is much bigger. Hence, I guess this system might not work too well - or will be quite resource hungry.

Side note: isn't it nice how copyright owners need to "submit" to Google?

eriky




msg:3478582
 7:55 am on Oct 16, 2007 (gmt 0)

Just commenting on the people that don't believe this can be done. I really believe it can be done by using for example artificial neural networks. You can make a fingerprint of a video by looking at characteristics that stay the same. For example, a video will have a certain fingerprint when you grayscale the entire video and only look at the color intensity at each moment of that video. You can do the same to a newly uploaded video. You create the fingerprint and start comparing against your database of fingerprints.
I don't know exactly how they do it but I am certain that it can be done. I think it is a genius solution as well. Why would a content owner object to people viewing his video if he get's his cut of the revenue. This might be the future of television.

tedster




msg:3478589
 8:06 am on Oct 16, 2007 (gmt 0)

Comment from eWeek - April 2007:

Google's not dumb. The G Poppa refuses to implement fingerprinting technology because doing so could implicate it in foreknowledge of the actual piracy. We'll filter, says Google, but only if you sign a licensing or partnership agreement. That way we're protected from lawsuits and you're protected from rampant piracy on our site.

Video Fingerprinting Won't Stop Piracy [googlewatch.eweek.com]


Quadrille




msg:3478594
 8:30 am on Oct 16, 2007 (gmt 0)

I was thinking the same thing - why don't they allow web site owners to upload a (legally binding) form saying, effectively "this content is mine and no one else's". Then calculate MD5 sums lets say for each paragraph longer than one sentence and put them on record.

1. It isn't their problem - they index the web, they don't publish it. YouTube is a publisher, which can and is being attacked by lawyers. There's no case against Google SE for listing an 'incorrect' web page.

2. There no legally binding form in existence to do that task. You could perjure yourself, Google could favour your page - then be successfully sued by the copyright owner; your perjury could not be used as defense - there's no legal framework for that.

3. The expense of manually reading every web page and checking for the declarations would break Google in a month.

4. Google are not the Web Police - and I'm amazed that anyone should want them to be - but they do need to police their own sites.

And there's more issues, I'll bet.

thecoalman




msg:3478617
 8:56 am on Oct 16, 2007 (gmt 0)

For example, a video will have a certain fingerprint when you grayscale the entire video and only look at the color intensity at each moment of that video.

If they were just trying to match high quality videos to other high quality videos I could agree. What needs to be understood is highly compressed video such as the ones uploaded to youtube are made up different types of frames. You have I, P and B frames. The only frame that is full frame is the I frame and they typically are spaced about 15 frames apart. The ones in between are made up of data of previous and subsequent frames, this why you see macroblocking with low bitrate video during high motion.

So the only reliable frame to match will be the I-frame. The location of this I-frame is going to vary according to the encoding method, length of video... So if 20 people upload the same video that was appropriated and prepared 20 different ways the I-frames will be not be in the same locations. Just to add the typical resolution of these videos is only 320x240 so there isn't a whole lot there to begin with. Highly compressed video depends a lot on tricking the eye into thinking you're seeing something.

As far as the color goes this too is going to vary depending on the method that it was appropriated, a video encoded from a DVD rip certainly is not going to be the same as one that was recorded from TV. Then you can throw into the mix the ability to crop, alter colors etc.

My argument is when all that is said in done considering the amount of video uploaded to youtube and the huge amount of processing power this is going to require (bear in mind video processing is about the most intense CPU activity you can throw at a processor) preventing a vast amount of false positives is going to be impossible.

Why would a content owner object to people viewing his video if he get's his cut of the revenue.

Because they only get to sell it once. The entertainment industry business model in the past has been to resell the same content over and over. If your middle age and like music you most likely have owned the same album first on vinyl, 8-track, cassette and finally CD. Until there is absolute secure method for them to protect digital content I wouldn't hold my breath waiting for that to happen.

thecoalman




msg:3478621
 9:00 am on Oct 16, 2007 (gmt 0)

Just wanted to add what I find real humorous is they want youtube to protect their content from being ditributed yet they cant do it themselves.

amznVibe




msg:3478654
 10:21 am on Oct 16, 2007 (gmt 0)

No way this system can match a clip segment out of a whole video, especially if the audio is altered or a logo is applied?

The CPU requirements for this must be massive?

Are content makers going to supply their entire inventory to Google? I think not.

cmendla




msg:3478655
 10:22 am on Oct 16, 2007 (gmt 0)

I know it's not exactly the same thing but..

I hope google's system works better than the simple copy protection system on my vhs to dvd deck. I'm trying to copy our videos to a dvd format. These are family videos and videos I shot for websites. In other words, it is unquestionably my work.

It gets really frustrating when the system stops and flashes a "COPY PROTECTION VIOLATION" during a scene of our family. Every time that happens I curse the hollywood misfits that have forced the electronics makers to err on the side of thinking that everything is stolen.

Also, My guess is that google's system is not geared to protecting little guys like me who post a unique video but rather for those who put up a clip of southpark or other major commercial work.

One final thought. As someone mentioned, there is ID software for music, Our wireless provider has a song id feature where you simply play a few seconds of a song and it ids it. Also, firewalls are able to filter out adult pictures by doing algos that determine how much skin is in a picture.

All in all, I don't expect google's anti piracy product to help the little guys much but I do expect that it will cause a lot of frustration when it labels legitimate content as copyright violation,

cg

thecoalman




msg:3478717
 11:45 am on Oct 16, 2007 (gmt 0)

It gets really frustrating when the system stops and flashes a "COPY PROTECTION VIOLATION" during a scene of our family.

That's a common problem for analog capture, commercial VHS tapes have macrovision protection. It's actually just an error in the tape, Macrovision patented this... Old VHS or Hi-8 tapes accumulate errors as they deteriorate, so your machine is seeing a regular error and thinks its copy protection. It's useless protection once converted to digital. It's destroyed during the encoding process or can simply be masked.

A TBC or "Video Clarifier" will alleviate this altogether, TBC if you have the money. It can do wonders for old tapes especially if they skewed, jumpy etc.

hughie




msg:3478774
 12:53 pm on Oct 16, 2007 (gmt 0)

What about current events, sports and other live tv? i know the Premier league here in the UK are rather annoyed with youtube for obvious reasons.

Are the premier league going to feed yotube live copies of every game and every goal from every angle...

joelgreen




msg:3478783
 1:03 pm on Oct 16, 2007 (gmt 0)

There is too many things you can do to alter the content, crop, blur... even the re-encoding process will produce different variants of the same video depending on the settings.

Lets say you select N spots on the video, and roughly track color (or maybe brightness/contrast/whatever) changes as video plays. If some high per cent (for example 80% or 90%) of the tracked points match original sample they can treat match successful. No need to compare frame to frame.

thecoalman




msg:3478836
 2:01 pm on Oct 16, 2007 (gmt 0)

Lets say you select N spots on the video, and roughly track color (or maybe brightness/contrast/whatever) changes as video plays. If some high per cent (for example 80% or 90%) of the tracked points match original sample they can treat match successful.

Let me give you a further example, you capture a movie off digital cable. Not sure of the exact format they use but it's most likely MPEG . It too has I, P, and B frames. The chances of the I frames being captured and encoded as a I-frames is like 1 in 15 so it's not very good. The new video you have is going to have what was a P or B frame as the new full I frame. the new P and B will be re-encoded. The new video is completely different that the old one.

As you can guess this degrades the video and is a destructive process, it's a common mistake noobies make when they are doing video. Compressed video should not be re-compressed if possible. If the skill of the person uploading isn't that great it will be all the worse.

Again my argument isn't that they can't do it, my argument is they can't do it without many false positives. To get it to youube it will be highly compressed which introduces yet even more degradation particularly during high motion and at a much lower resolution than the original.

It's just like saving the same .jpg over and over again. Atifacts start to pop out left and right. The atifacts produced by 20 different people using 20 different methods are going to be completely different for each video.

hughie




msg:3478890
 2:46 pm on Oct 16, 2007 (gmt 0)

don't forget here googles "10,000 strong " army of people.

Are they using this system to flag possibles, which then go for manual review?

gibbergibber




msg:3478983
 4:45 pm on Oct 16, 2007 (gmt 0)

The technical problem isn't so much the comparison mechanism, it's where the heck you get the reference material from.

Even just listing the titles of a single film company's back catalogue is a considerable task. To do reference images for a majority of films from a majority of film and TV companies would be even more huge.

Like someone said above, this is just a figleaf to satisfy those companies who are particularly annoyed by a specific bit of video material being pirated.

hutcheson




msg:3479042
 5:57 pm on Oct 16, 2007 (gmt 0)

It's more than just a figleaf. It throws the ball back into the Luddite court.

Obviously, copyright infringement isn't theft. But, like theft, there must be a specific asset which has been used in an unlicensed manner. If a Walmart executive went to the police station and accuses your local library of containing stolen materials, you wouldn't want the police to lock down the library. You'd expect Walmart to identify the specific books being stolen. After all, the police don't even have a list of the Walmart inventory, to know what MIGHT have been taken.

No, Walmart absolutely has to identify the goods, or there's no possible case outside of Kangaroo Court.

Well, that's theft. But other things are no different. If you accuse someone of murder, you'd better be able to provide a corpse (or at least, a name of a missing person.) And in the infringement racket, it's no different. Google has no earthly idea what all trash has been shown in newsreels or movies, or broadcast on TV or cable, or sold on VCR or DVD.

So, Google has provided a tool. We'll see if it works, technologically. (I'm betting it works well enough in the lab to be persuasive, but not well enough to make any practical difference, other than keeping the Luddite minions busy taking useless fingerprints of material nobody ever thought to copy.)

I think it'll work socially: it'll provide a way for Google to dump the blame on the people whose responsibility it is: the content-barons themselves. Because it this doesn't work, Google can challenge the Luddites to prove something better can exist. Prove, that is, by building it, of course.

That should be fun to watch.

celerityfm




msg:3479057
 6:20 pm on Oct 16, 2007 (gmt 0)

My biggest question is what are the possibilities for abuse of this system? There have already been cases where companies have filed inappropriate or illegal DMCA takedown notices with YouTube to remove videos they had a problem with. With the DMCA you can file counter-takedown notices (and there could be fines for violators), but what will this system have in place to protect from abuse?

After reviewing the Youtube page about this [youtube.com] I feel that the way they describe the system it would be possible for anyone to just rip an FLV file from a YouTube page and then simply upload it to YouTube's "copyright protection" area and then ask YouTube to delete any videos that are similar in nature.

Seems like this system might become a new way to block political speech, whistleblowers or just about any speech that someone else doesn't like. Google's move here to help copyright holders could end up being at the expense of the "anything-goes" free-speech nature of YouTube :(

crobb305




msg:3479908
 4:13 pm on Oct 17, 2007 (gmt 0)

How can they detect duplicate videos, but they can't detect duplicated/stolen content that comes subsequent to the original?

Quadrille




msg:3479954
 4:45 pm on Oct 17, 2007 (gmt 0)

They can.

But first use does not necessarily mean ownership; and anyway, Google is not the web police, and most of us would not want them to be, I suspect.

And in the YouTube plan, owners are going to have to co-operate to make it work.

crobb305




msg:3479983
 5:24 pm on Oct 17, 2007 (gmt 0)

Google is not the web police

No one suggested that for heaven's sake. But if Google were able to identifiy originals, first-to-use (which is usually the original on the internet unless someone is scanning printed documents), etc, then there would be no problem of proxies and blantant website duplicates outranking the originals. If they can address the duplicate content issue with video, why can't they do the same for content search? Both are equally important.

C

This 35 message thread spans 2 pages: 35 ( [1] 2 > >
Global Options:
 top home search open messages active posts  
 

Home / Forums Index / Google / Google SEO News and Discussion
rss feed

All trademarks and copyrights held by respective owners. Member comments are owned by the poster.
Home ¦ Free Tools ¦ Terms of Service ¦ Privacy Policy ¦ Report Problem ¦ About ¦ Library ¦ Newsletter
WebmasterWorld is a Developer Shed Community owned by Jim Boykin.
© Webmaster World 1996-2014 all rights reserved