Find Image Plagiarism with FeelImage

feelimage.jpgNew photo search engine FeelImage aims to make it easier to search for photographs by their “feeling”, in particular their color and time.

Though it might seem like something of a novelty, the ability for users to search for “purple flower” or “red car” might have an unintended side effect, enabling photographers and artists to better search for copies of their work online.

With most current photo searche engines based solely upon tagging and words around the image, there are very few techniques to effectively search for duplicates of an image on the Web. Searching for the file name, for example, is one possibility, but it is easily defeated by simply renaming the file before re-uploading it.

Though FeelImage, in its current incarnation, is only a minor step forward, it is already being used by some artists, including the Ripped Art Task Force, to track down copied images.

Other visual artists may find it just as useful.

Usage

The idea behind using FeelImage is pretty straightforward. If you have an image that is often stolen or you suspect might be, simply search for the color (or time) and a description of what is in the image. FeelImage, in a few moments, will pull up a list of photos similar to the one you describe. From there, it is easy to browse and look for any exact copies of your image.

It is important to note though that, when using FeelImage, that the color detection is automated and, thus, not perfect. If you have a picture of a red car that is set against a blue sky, FeelImage will likely read most of the image as being blue and it might not show up well on a search for “Red Car” but will show up higher on “Blue Car”.

Furthermore, FeelImage works best with photos and images that are of one subject and, for the most part, one color. The more complicated the image is, the less likely FeelImage will accurately describe it.

Of course, these are not FeelImage’s only limitations. The service is very new and it is important to be aware that it is still a work in progress. Thus, its search comes with a few caveats.

Limitations

FeelImage, as its logo indicates, is only an alpha release. It is not feature complete and has many bugs in it.

The greatest limitation is that FeelImage only searches for photos on three photo services, Photozou, Zorg and Flickr.

Though Flickr is a very popular photo hosting service and is used to display images on many different sites, it is not the only image hosting service on the Web by any stretch. Photos hosted by Imagecshack, Photobucket and others will go undetected.

Second, the object information seems to be based upon the tags that the user supplies with the photos. Though the color detection seems to work automatically, perhaps to a fault, FeelImage has to look at the tags to determine what it is a photo of. Though creating a system to do that automatically would be virtually impossible, the reliance on user-submitted input, in this case a plagiarist, is a definite weakness.

Finally, since the service is still in a pre-release phase, the results are not as accurate as they could be. It is clear that the developers are still working on the software that runs the search engine as, in some searches, the first few pages of results are almost completely irrelevant. Some searches, however, work better than others and it is important to play around with modified searches to see what results you get.

Still, these limitations should not stop artists from at least experimenting with FeelImage. Not only can it be used to detect plagiarism, but its intended use, finding similar images based upon color and time, is a very interesting feature that produces some unexpected results.

It is a great tool and it will be interesting to see how it grows.

Conclusions

Image plagiarism is easily the hardest to detect. Though tools are available to monitor such infringements, they are generally out of the reach of your average blogger or amateur artist.

FeelImage represents only a small step forward in detecting image plagiarism but it is an important one. It finally gets through the use of text and tags and looks into the image itself for clues. This can help narrow down the search results and increase the likelihood that the image you want is at the top of the results.

Despite its severe limitations, at least at this time, it is an important site for artists, photographers and others who post images to be aware of.

Not only is it a powerful tool today, but it could easily grow into something much greater and provide an even bigger boost to artists in the future. It will definitely be a service to follow.

59 comments
JB
JB

Cybele, I don't want to count them out either. I really like the idea behind what they are trying to do but they need to obey the same rules as everyone else and I agree that they were doing some things that were wrong. I hope they can go back and fix those issues. I'd like to see them apply for an actual API key and go that way, it'd work better in the long run. It is a neat service, this was, in truth, probably more of an oversight than anything. I hope they can fix it and get back on track.

JB
JB

Cybele,
I don't want to count them out either. I really like the idea behind what they are trying to do but they need to obey the same rules as everyone else and I agree that they were doing some things that were wrong.
I hope they can go back and fix those issues. I'd like to see them apply for an actual API key and go that way, it'd work better in the long run.
It is a neat service, this was, in truth, probably more of an oversight than anything. I hope they can fix it and get back on track.

JB
JB

Cybele,

I don't want to count them out either. I really like the idea behind what they are trying to do but they need to obey the same rules as everyone else and I agree that they were doing some things that were wrong.

I hope they can go back and fix those issues. I'd like to see them apply for an actual API key and go that way, it'd work better in the long run.

It is a neat service, this was, in truth, probably more of an oversight than anything. I hope they can fix it and get back on track.

JB
JB

Cybele,

I don't want to count them out either. I really like the idea behind what they are trying to do but they need to obey the same rules as everyone else and I agree that they were doing some things that were wrong.

I hope they can go back and fix those issues. I'd like to see them apply for an actual API key and go that way, it'd work better in the long run.

It is a neat service, this was, in truth, probably more of an oversight than anything. I hope they can fix it and get back on track.

cybele
cybele

I'm not sure we should count them out, they'll just need to query Flickr in a manner that's consistent with the opt in/op out preferences by the photographers.

If they can actually put the correct license (whether CC or (c)) on the page, not create an embedding code that violates Flickr policies and of course only pull data on photos that have opted into the outside searches it would be a great tool.

cybele
cybele

I'm not sure we should count them out, they'll just need to query Flickr in a manner that's consistent with the opt in/op out preferences by the photographers.

If they can actually put the correct license (whether CC or (c)) on the page, not create an embedding code that violates Flickr policies and of course only pull data on photos that have opted into the outside searches it would be a great tool.

cybele
cybele

I'm not sure we should count them out, they'll just need to query Flickr in a manner that's consistent with the opt in/op out preferences by the photographers.
If they can actually put the correct license (whether CC or (c)) on the page, not create an embedding code that violates Flickr policies and of course only pull data on photos that have opted into the outside searches it would be a great tool.

cybele
cybele

I'm not sure we should count them out, they'll just need to query Flickr in a manner that's consistent with the opt in/op out preferences by the photographers. If they can actually put the correct license (whether CC or (c)) on the page, not create an embedding code that violates Flickr policies and of course only pull data on photos that have opted into the outside searches it would be a great tool.

JB
JB

Cybele: If you haven't already, you might want to check out yesterday's post on the subject. It appears that FeelImage is no longer indexing Flickr. It appears that they have been blocked from doing so.

It is sad that they could not get a handle on these copyright issues, it is, potentially an exciting.

However, you are right. There is a lot of reason to not trust feel Image. I had just assumed they were using the Flickr API, which would have been the easiest way, but it appears I was wrong.

Jess: Looks interesting but, sadly, is also infringing itself in some ways. A terrible shame.

Thank you for the explanation on how Flickr handles EXIF data.

JB
JB

Cybele: If you haven't already, you might want to check out yesterday's post on the subject. It appears that FeelImage is no longer indexing Flickr. It appears that they have been blocked from doing so.
It is sad that they could not get a handle on these copyright issues, it is, potentially an exciting.
However, you are right. There is a lot of reason to not trust feel Image. I had just assumed they were using the Flickr API, which would have been the easiest way, but it appears I was wrong.
Jess: Looks interesting but, sadly, is also infringing itself in some ways. A terrible shame.
Thank you for the explanation on how Flickr handles EXIF data.

JB
JB

Cybele: If you haven't already, you might want to check out yesterday's post on the subject. It appears that FeelImage is no longer indexing Flickr. It appears that they have been blocked from doing so. It is sad that they could not get a handle on these copyright issues, it is, potentially an exciting. However, you are right. There is a lot of reason to not trust feel Image. I had just assumed they were using the Flickr API, which would have been the easiest way, but it appears I was wrong. Jess: Looks interesting but, sadly, is also infringing itself in some ways. A terrible shame. Thank you for the explanation on how Flickr handles EXIF data.

JB
JB

Cybele: If you haven't already, you might want to check out yesterday's post on the subject. It appears that FeelImage is no longer indexing Flickr. It appears that they have been blocked from doing so.

It is sad that they could not get a handle on these copyright issues, it is, potentially an exciting.

However, you are right. There is a lot of reason to not trust feel Image. I had just assumed they were using the Flickr API, which would have been the easiest way, but it appears I was wrong.

Jess: Looks interesting but, sadly, is also infringing itself in some ways. A terrible shame.

Thank you for the explanation on how Flickr handles EXIF data.

JB
JB

Cybele,I don't want to count them out either. I really like the idea behind what they are trying to do but they need to obey the same rules as everyone else and I agree that they were doing some things that were wrong.I hope they can go back and fix those issues. I'd like to see them apply for an actual API key and go that way, it'd work better in the long run.It is a neat service, this was, in truth, probably more of an oversight than anything. I hope they can fix it and get back on track.

JB
JB

Cybele,I don't want to count them out either. I really like the idea behind what they are trying to do but they need to obey the same rules as everyone else and I agree that they were doing some things that were wrong.I hope they can go back and fix those issues. I'd like to see them apply for an actual API key and go that way, it'd work better in the long run.It is a neat service, this was, in truth, probably more of an oversight than anything. I hope they can fix it and get back on track.

cybele
cybele

I'm not sure we should count them out, they'll just need to query Flickr in a manner that's consistent with the opt in/op out preferences by the photographers. If they can actually put the correct license (whether CC or (c)) on the page, not create an embedding code that violates Flickr policies and of course only pull data on photos that have opted into the outside searches it would be a great tool.

cybele
cybele

I'm not sure we should count them out, they'll just need to query Flickr in a manner that's consistent with the opt in/op out preferences by the photographers. If they can actually put the correct license (whether CC or (c)) on the page, not create an embedding code that violates Flickr policies and of course only pull data on photos that have opted into the outside searches it would be a great tool.

JB
JB

Cybele: If you haven't already, you might want to check out yesterday's post on the subject. It appears that FeelImage is no longer indexing Flickr. It appears that they have been blocked from doing so.It is sad that they could not get a handle on these copyright issues, it is, potentially an exciting. However, you are right. There is a lot of reason to not trust feel Image. I had just assumed they were using the Flickr API, which would have been the easiest way, but it appears I was wrong.Jess: Looks interesting but, sadly, is also infringing itself in some ways. A terrible shame. Thank you for the explanation on how Flickr handles EXIF data.

JB
JB

Cybele: If you haven't already, you might want to check out yesterday's post on the subject. It appears that FeelImage is no longer indexing Flickr. It appears that they have been blocked from doing so.It is sad that they could not get a handle on these copyright issues, it is, potentially an exciting. However, you are right. There is a lot of reason to not trust feel Image. I had just assumed they were using the Flickr API, which would have been the easiest way, but it appears I was wrong.Jess: Looks interesting but, sadly, is also infringing itself in some ways. A terrible shame. Thank you for the explanation on how Flickr handles EXIF data.

Jess
Jess

Looks interesting, great post as always! :)

Jess
Jess

Looks interesting, great post as always! :)

Jess
Jess

Looks interesting, great post as always! :)

cybele
cybele

On flickr users have the ability to hide exif data and their original photos (flickr creates different sizes when you upload and users have the option to make some or all available to the public).

Here's the forum topic I started on the subject and it does not appear to be within Flickr's TOU because they provide a code for people to hotlink to an image on Flickr, but not link back to the actual image without visiting FeelImage first.

http://www.flickr.com/help/forum/49579/

Basically, I use flickr and I've opted to hide my photos from searches outside flickr (yahoo, google, etc.) and everyone else seems to honor that. (My images are not shown in tag search RSS feeds from the site either ... only people who have specifically subscribed to my images or if I've shared an image with a public group).

In that forum thread other users have pointed out that FeelImage is not representing the copyright on the photos accurately either.

cybele
cybele

On flickr users have the ability to hide exif data and their original photos (flickr creates different sizes when you upload and users have the option to make some or all available to the public).

Here's the forum topic I started on the subject and it does not appear to be within Flickr's TOU because they provide a code for people to hotlink to an image on Flickr, but not link back to the actual image without visiting FeelImage first.

http://www.flickr.com/help/forum/49579/

Basically, I use flickr and I've opted to hide my photos from searches outside flickr (yahoo, google, etc.) and everyone else seems to honor that. (My images are not shown in tag search RSS feeds from the site either ... only people who have specifically subscribed to my images or if I've shared an image with a public group).

In that forum thread other users have pointed out that FeelImage is not representing the copyright on the photos accurately either.

cybele
cybele

On flickr users have the ability to hide exif data and their original photos (flickr creates different sizes when you upload and users have the option to make some or all available to the public).
Here's the forum topic I started on the subject and it does not appear to be within Flickr's TOU because they provide a code for people to hotlink to an image on Flickr, but not link back to the actual image without visiting FeelImage first.
http://www.flickr.com/help/forum/49579/
Basically, I use flickr and I've opted to hide my photos from searches outside flickr (yahoo, google, etc.) and everyone else seems to honor that. (My images are not shown in tag search RSS feeds from the site either ... only people who have specifically subscribed to my images or if I've shared an image with a public group).
In that forum thread other users have pointed out that FeelImage is not representing the copyright on the photos accurately either.

cybele
cybele

On flickr users have the ability to hide exif data and their original photos (flickr creates different sizes when you upload and users have the option to make some or all available to the public). Here's the forum topic I started on the subject and it does not appear to be within Flickr's TOU because they provide a code for people to hotlink to an image on Flickr, but not link back to the actual image without visiting FeelImage first. http://www.flickr.com/help/forum/49579/ Basically, I use flickr and I've opted to hide my photos from searches outside flickr (yahoo, google, etc.) and everyone else seems to honor that. (My images are not shown in tag search RSS feeds from the site either ... only people who have specifically subscribed to my images or if I've shared an image with a public group). In that forum thread other users have pointed out that FeelImage is not representing the copyright on the photos accurately either.

cybele
cybele

On flickr users have the ability to hide exif data and their original photos (flickr creates different sizes when you upload and users have the option to make some or all available to the public). Here's the forum topic I started on the subject and it does not appear to be within Flickr's TOU because they provide a code for people to hotlink to an image on Flickr, but not link back to the actual image without visiting FeelImage first. http://www.flickr.com/help/forum/49579/Basically, I use flickr and I've opted to hide my photos from searches outside flickr (yahoo, google, etc.) and everyone else seems to honor that. (My images are not shown in tag search RSS feeds from the site either ... only people who have specifically subscribed to my images or if I've shared an image with a public group). In that forum thread other users have pointed out that FeelImage is not representing the copyright on the photos accurately either.

JB
JB

Cybele,

That is an interesting point. Granted, I'm not familiar with Flickr's TOU (I'm not a member myself and I don't link to Flickr images) but that is an issue that needs to be resolved.

However, it seems likely that the service is being cut some slack. Given all of the media coverage, I'm almost certain that Yahoo is aware of the site. Perhaps its background in education (it was created in part by a University) has something to do with.

I doubt they could have gotten so far violating Flickr's TOU without permission.

On one side note, I'm not sure there is such a thing as "private" EXIF data because that is merely data embedded in the image, any program capable of reading EXIF data can extract it. Unless you encrypted it, there's no way to make such data private.

However, I will contact Flickr about this and see what they have to say. This is very interesting and very worrisome.

JB
JB

Cybele,

That is an interesting point. Granted, I'm not familiar with Flickr's TOU (I'm not a member myself and I don't link to Flickr images) but that is an issue that needs to be resolved.

However, it seems likely that the service is being cut some slack. Given all of the media coverage, I'm almost certain that Yahoo is aware of the site. Perhaps its background in education (it was created in part by a University) has something to do with.

I doubt they could have gotten so far violating Flickr's TOU without permission.

On one side note, I'm not sure there is such a thing as "private" EXIF data because that is merely data embedded in the image, any program capable of reading EXIF data can extract it. Unless you encrypted it, there's no way to make such data private.

However, I will contact Flickr about this and see what they have to say. This is very interesting and very worrisome.

JB
JB

Cybele,
That is an interesting point. Granted, I'm not familiar with Flickr's TOU (I'm not a member myself and I don't link to Flickr images) but that is an issue that needs to be resolved.
However, it seems likely that the service is being cut some slack. Given all of the media coverage, I'm almost certain that Yahoo is aware of the site. Perhaps its background in education (it was created in part by a University) has something to do with.
I doubt they could have gotten so far violating Flickr's TOU without permission.
On one side note, I'm not sure there is such a thing as "private" EXIF data because that is merely data embedded in the image, any program capable of reading EXIF data can extract it. Unless you encrypted it, there's no way to make such data private.
However, I will contact Flickr about this and see what they have to say. This is very interesting and very worrisome.

JB
JB

Cybele, That is an interesting point. Granted, I'm not familiar with Flickr's TOU (I'm not a member myself and I don't link to Flickr images) but that is an issue that needs to be resolved. However, it seems likely that the service is being cut some slack. Given all of the media coverage, I'm almost certain that Yahoo is aware of the site. Perhaps its background in education (it was created in part by a University) has something to do with. I doubt they could have gotten so far violating Flickr's TOU without permission. On one side note, I'm not sure there is such a thing as "private" EXIF data because that is merely data embedded in the image, any program capable of reading EXIF data can extract it. Unless you encrypted it, there's no way to make such data private. However, I will contact Flickr about this and see what they have to say. This is very interesting and very worrisome.

cybele
cybele

FeelImage is not quite what it seems. It draws largely from photos hosted on Flickr and seems to be violating Flickr's linkback policies (and doesn't seem to include the copyright notices attached to the photos, though it does list the CC license). There's a little "embed this photo" link at the bottom of the photo info page ... when you embed that photo it's hosted at flickr, but clicking on the photo takes you to FeelImage (a violation of Flickr TOU).

Further, it appears to be displaying private photos, private EXIF data and photos that are not supposed to be available to API searches through flickr.

cybele
cybele

FeelImage is not quite what it seems. It draws largely from photos hosted on Flickr and seems to be violating Flickr's linkback policies (and doesn't seem to include the copyright notices attached to the photos, though it does list the CC license). There's a little "embed this photo" link at the bottom of the photo info page ... when you embed that photo it's hosted at flickr, but clicking on the photo takes you to FeelImage (a violation of Flickr TOU).

Further, it appears to be displaying private photos, private EXIF data and photos that are not supposed to be available to API searches through flickr.

cybele
cybele

FeelImage is not quite what it seems. It draws largely from photos hosted on Flickr and seems to be violating Flickr's linkback policies (and doesn't seem to include the copyright notices attached to the photos, though it does list the CC license). There's a little "embed this photo" link at the bottom of the photo info page ... when you embed that photo it's hosted at flickr, but clicking on the photo takes you to FeelImage (a violation of Flickr TOU). Further, it appears to be displaying private photos, private EXIF data and photos that are not supposed to be available to API searches through flickr.

cybele
cybele

FeelImage is not quite what it seems. It draws largely from photos hosted on Flickr and seems to be violating Flickr's linkback policies (and doesn't seem to include the copyright notices attached to the photos, though it does list the CC license). There's a little "embed this photo" link at the bottom of the photo info page ... when you embed that photo it's hosted at flickr, but clicking on the photo takes you to FeelImage (a violation of Flickr TOU).
Further, it appears to be displaying private photos, private EXIF data and photos that are not supposed to be available to API searches through flickr.

JB
JB

Andrew,

Like I said, the accuracy is mixed, give it a shot and let me know what you find out!

JB
JB

Andrew,

Like I said, the accuracy is mixed, give it a shot and let me know what you find out!

JB
JB

Andrew, Like I said, the accuracy is mixed, give it a shot and let me know what you find out!

JB
JB

Andrew,
Like I said, the accuracy is mixed, give it a shot and let me know what you find out!

Andrew
Andrew

Great idea, I wonder how accurate this will be though...

Andrew
Andrew

Great idea, I wonder how accurate this will be though...

Andrew
Andrew

Great idea, I wonder how accurate this will be though...

JB
JB

Cybele,That is an interesting point. Granted, I'm not familiar with Flickr's TOU (I'm not a member myself and I don't link to Flickr images) but that is an issue that needs to be resolved.However, it seems likely that the service is being cut some slack. Given all of the media coverage, I'm almost certain that Yahoo is aware of the site. Perhaps its background in education (it was created in part by a University) has something to do with.I doubt they could have gotten so far violating Flickr's TOU without permission. On one side note, I'm not sure there is such a thing as "private" EXIF data because that is merely data embedded in the image, any program capable of reading EXIF data can extract it. Unless you encrypted it, there's no way to make such data private. However, I will contact Flickr about this and see what they have to say. This is very interesting and very worrisome.

JB
JB

Cybele,That is an interesting point. Granted, I'm not familiar with Flickr's TOU (I'm not a member myself and I don't link to Flickr images) but that is an issue that needs to be resolved.However, it seems likely that the service is being cut some slack. Given all of the media coverage, I'm almost certain that Yahoo is aware of the site. Perhaps its background in education (it was created in part by a University) has something to do with.I doubt they could have gotten so far violating Flickr's TOU without permission. On one side note, I'm not sure there is such a thing as "private" EXIF data because that is merely data embedded in the image, any program capable of reading EXIF data can extract it. Unless you encrypted it, there's no way to make such data private. However, I will contact Flickr about this and see what they have to say. This is very interesting and very worrisome.

JB
JB

Stay at Home Dad, The effectiveness of it really depends on the image that you're looking for. If it's a one-color image (though not black and white) that has just one subject, it can be pretty effective. If you're looking for something much more complicated, it won't work as well. Judging from my interactions with the Ripped Art Task Force, it is already showing some impressive results. In fact, on their first search, which was just for playing around, they found a ripped photo. Its effectiveness is limited but in cases where the image plays to its strengths, it works pretty well. Hope that helps!

JB
JB

Stay at Home Dad,
The effectiveness of it really depends on the image that you're looking for. If it's a one-color image (though not black and white) that has just one subject, it can be pretty effective. If you're looking for something much more complicated, it won't work as well.
Judging from my interactions with the Ripped Art Task Force, it is already showing some impressive results. In fact, on their first search, which was just for playing around, they found a ripped photo.
Its effectiveness is limited but in cases where the image plays to its strengths, it works pretty well.
Hope that helps!

JB
JB

Stay at Home Dad,

The effectiveness of it really depends on the image that you're looking for. If it's a one-color image (though not black and white) that has just one subject, it can be pretty effective. If you're looking for something much more complicated, it won't work as well.

Judging from my interactions with the Ripped Art Task Force, it is already showing some impressive results. In fact, on their first search, which was just for playing around, they found a ripped photo.

Its effectiveness is limited but in cases where the image plays to its strengths, it works pretty well.

Hope that helps!

JB
JB

Stay at Home Dad,

The effectiveness of it really depends on the image that you're looking for. If it's a one-color image (though not black and white) that has just one subject, it can be pretty effective. If you're looking for something much more complicated, it won't work as well.

Judging from my interactions with the Ripped Art Task Force, it is already showing some impressive results. In fact, on their first search, which was just for playing around, they found a ripped photo.

Its effectiveness is limited but in cases where the image plays to its strengths, it works pretty well.

Hope that helps!

Stay At Home Dad, Geek Style
Stay At Home Dad, Geek Style

Any idea how effective this system is in its current form? I think it is something that could make the process of finding stolen images much easier, but there must be literally hundreds of ways around it. DevDad

Stay At Home Dad, Geek Style
Stay At Home Dad, Geek Style

Any idea how effective this system is in its current form? I think it is something that could make the process of finding stolen images much easier, but there must be literally hundreds of ways around it.

DevDad

Stay At Home Dad, Geek Style
Stay At Home Dad, Geek Style

Any idea how effective this system is in its current form? I think it is something that could make the process of finding stolen images much easier, but there must be literally hundreds of ways around it.

DevDad

Trackbacks

  1. [...] please consider subscribing to my RSS feed. Thank you for visiting!Image search engine FeelImage, previously discussed here, has stopped displaying Flickr results in its [...]

philosophical