Every few days I get a message from users all over the world, with more or less the same question: I have a photo of someone I am looking for, can you help me find this person on the Internet? (This is usually referred to “reverse image search“, or “people search” with photograph images, or “similar image searching“).
Unfortunately, I can’t. I’d need to have access to all images of all people of this world to use face recognition on them. Some Intelligence Service might have such a database. I don’t.
You need to have access to a real huge collection of images, and you need very strong computing power. That can only be done by companies with huge resources. Therefore, for some time it was only Google with such a capability. Their image search is not using face recognition (afaik), but their scale invariant feature technology works nice on face images, too. But now it is no longer Google alone to offer such a service! So here is a list of search engines that can compare images within a reasonably large database (please send feedback if you know more):
- Google Image Search Just upload a photo to Google’s Image Search, and it will find similar images. With Google owning probably the largest database of images on the web, chances are high that you find that (or at least a similar) photo.
- ImageRaider Currently uses Google, Bing and Yandex to do a reverse image search by making use of the IncandescentAPI. Very helpful site for people using a mobile device and therefore cannot easily use the links above.
- ImageBrief Especially made for photographers. Also helps finding brand images. (Click on the camera symbol within the search field)
- PimEyes Uses real facial recognition algorithms to compare a submitted photo with images of about 10 million people. The database obviously contains many freely available images from the Web, including Wikipedia. Results are quite good even for rotated, low-res faces.
- Berify Stolen Image Search Last but surely not least I’d like to draw your attention to this new reverse image search engine (which seems to be a spin-off of social catfish). Officially made to search for “stolen” images, it is quite powerful and different from all the others above. Their algorithm to match pictures, in my opinion, seems to use more shape and colour information than the other engines. I mean, if you reverse search a picture of a blonde woman with green eyes and blue shirt, you will get pictures like that with that engine. The other engines above might give you totally different results of a red-haired guy with green shirt. So whatever those guys at Berify did, I think it is worth to try out this image search catfish (use the free trial version).
Very often people are sending pictures from someone who contacted them, by some chat community or partnership platform. They want to know whether this person is for real, or whether it is a so-called “scammer”, using pictures of someone else. This seems to be a big problem called “romance scam“. A few websites try to help users by searching for images of well-known scammers:
- Scam Digger Pic Search Beta version using images from Romancescam forums.
Now that this topic is not fully related to face detection, I created a page especially for listing all related tools and resources. If you want to know more about reverse image search, please visit Reverse Image Search Resources!
Other interesting demos using face detection / face finders technology:
- CelebsLike.Me from Microsoft. Which Oscar nominee do you look like? A face matching app from Microsoft using their new deep-learning technology. Upload a photo and you see which 2016 Oscar nominee looks like the person on that pic.
- How old do I look? from Microsoft. You can upload a photo with any amount of faces on it, and the site will answer with all found faces marked, together with an age estimation of each person.
Don’t be too shocked if your estimated age is far off reality (+- 15 years accuracy is my rough guess), but the face detector seems to be highly accurate. Microsoft uses its Project Oxford web service for doing that, so you might want to check that out.
- Real-time Face Detector 2.0 from Alexander Telnykh. Detection of faces in frontal angle on digital photos and in videostreams, detection of faces in a tilt perspective. Allows front objects between ±15° from horizontal, detection of faces in profile perspective, demo of Eyes Localization, demo of Face Tracking, demo of skin-color filtering.
- Submit face images to the University Illinois online face finder