Have you ever wondered whether photos you take are used elsewhere online? Flickr has recently announced a partnership with a website which really grabbed my interest.
Pro members of Flickr have been given free access to Pixsy, which is a service designed to fight online photo theft. After reading a bit more about it, I signed up, and Pixsy started scanning my Flickr timeline and reporting back. I need to mention that this is not an endorsement or paid content for Flickr or Pixsy and is purely my own, initial impression of a photography service I found really interesting.
As a hobbyist photographer who has never received any royalties for any images, I have wondered what I’d do if I saw any website using my photos for profit. I’ve seen a few of my photos on skyscrapercity.com before, in threads where people post photos they like from places they live or visit, generally with direct links and/or credit. And stats for Flickr pro often include referrals which show sites where a photo of mine has been linked from. By and large it isn’t anything to be bothered about and I find it interesting that anyone would want to use one of my photos.
Back to Pixsy, however, and just a few minutes into the scan, the results start to come in: 730 images found. I dive in to the stats…
Immediately it identifies photos that I reposted myself. They’re my additions to Flickr groups and plenty of photos from this site. That’s a decent start, I guess, since it at least shows the service is picking up known matching images. These domains can be added to an approved list so they don’t pop up again.
After the initial surprise of seeing over 700 images found, I start digging into a few in more detail. First off, the service seems to struggle a bit with sunsets. Two photos in particular matched with sunset images from other sites which were not mine and this was immediately obvious, but something about the images seemed to fool the algorithm the site uses. I’m able to note these as mismatches which hopefully feeds back into the algorithm to improve future comparisons.
The close calls
Results can be filtered by accuracy for high, medium and low confidence in the matches and I started with the ‘highs’. There was a photo I took from Senate Square in Helsinki of the cathedral which popped up with a match of a monochrome photo from elsewhere. The crop, angle, lighting and everything I could immediately compare looked like it was a copy of my picture which had been desaturated. It was uncanny, but for a tiny spot of light I eventually saw. Ultimately it meant that someone else in a popular area in a big city had stood in the same place, pointed their camera in the same direction and taken almost exactly the same photo. It’s going to happen in a busy place.
The view from Hallgrimskirkja in Reykjavik was another which popped up with a few dozen similar images and for a very similar reason as the Helsinki photo. Then view from the top of the church is through such a tiny window, and lots of people line up their shot up with the street below, so it’s expected that something like this will result in many similar photos.
There was a random doorway from Basel, where the result showed the same doorway, but a photo taken by another person with a different crop. Multiple matches came up from the Duomo in Milan, none of which were actually mine. And there was a particular view from a spot in Zurich which came back as a match, but an incorrect one.
Many of these could have been a copied photo with a different crop and some colour changes. For all that these were not my photos, it is interesting to see an algorithm correctly identify photos taken from the same spot which match so closely that I couldn’t tell the difference immediately. These can be noted as “not my image” to remove them from the results.
In amongst the results there are hundreds of matches and it tends to be with a bit of theme. Around 20 photos have been used multiple times across multiple sites and it looks like, once used, these spread. Many of these sites appear to peddle nothing of interest. They’re from URLs that suggest they’re image scrapers or aggregators, one of which I clicked on which tried to kick off some malicious scripts in the browser. A few other sites I couldn’t even determine the point of.
After working through a few of these, it became a bit dull. I was looking to find photos that had been used in “proper” websites.
The “interesting” results
While picking through a few results, I was taking the source URL and pasting it into a browser with scripting turned off. One of these was a photo of a monastery taken in Kiev… being used on a site which allows users to add a picture of a stripper in a chosen pose on top of selected images, and the monastery was a stock image on the site. Rule 34 of the internet, anyone?
After filtering through the false positives, casting an eye over very similar photos and working through some seemingly weird and whacky websites, I get to a handful of results which do grab my attention. There’s a filter on the searches which identifies commercial sites, and adding this with the ‘high confidence’ filter quickly gets down to the results which would need the attention of a pro photographer or anyone who sells their material.
There’s a holiday website, using a photo of mine from Kiev to advertise the city.
There’s a site advertising Memphis, which uses a night shot I took a few years ago.
There’s a commercial property for sale in the UK which is using my image as its main photo on the estate agent listing.
There’s a blog post on the Northern Lights which uses a picture I took in Kangerlussuaq (not even a particularly good one!)
There are multiple desktop wallpaper websites using some of my landscape shots, making these available for download.
In the case of the Memphis photo, the website provides a credit and links back to Flickr so if anything it’s a bit of flattery. The blog post is gathering together aurora photos the author likes – some are credited and others aren’t. But the holiday website, the wallpaper websites, they’re making money from these through sales and/or advertising and it’s entirely unfair to be using anyone else’s photos in these circumstances. In the case of the estate agency stealing a photo to use on their listing, that’s going too far. A nice feature of the link up between Flickr and Pixsy is ten free “take down” notices, so one of these is getting used.
Interestingly, none of the results returned my Instagram account, which includes a lot of identical content to my Flickr feed. Instagram could be an easy and well-used source for copyright infringement so hopefully the scan will pick up this at some point. Similarly, I know of other photos posted on skyscrapercity.com (with credit) which haven’t yet been picked up.
As I type, the scan is still going (the web is a big place, I guess) so it will be interesting to see what else Pixsy highlights. For pro photographers and people making a living through selling their images, this kind of service must be fantastic. For a hobbyist like me, it’s more something I’m interested in rather than to use for on-going monitoring and multiple take down requests. Now, back to the latest scan results…