A cousin of mine was doing some Thorp research, and found several unidentified photos that were in an album with some AJ Thorp photos and was wondering if there is a good facial recognition software that may be helpful in identifying the men in the photos? In other words, could we teach such software with identified photos of men of various ages that are identified, and use that to get suggestions for the unidentified photos? If not, what might be some other tools we might use?
Facial recognition software is improving all the time so, yes, we could teach it to help your search.
If you want a package that will tick away and compare pictures, looking for matches with a set of photographs, there are several on the market. None of them cheap. You can find them with your favourite search engine.
If you want something to try a quick check on a few pictures, there are a couple of services on the web:
Pictriev and the Microsoft site Twins or Not will both accept two photographs and then score a probability that the photographs are of the same person. These may help in what you seek but are not a perfect answer. Both sites are improving as more use is made of them - giving them the opportunity to learn.
In the future another possibility may arrive: Image recognition is one of the so-called superfunctions of the Wolfram language. It does not yet do facial recognition in the way you seek but it may do as the language grows.
I don't use Picasa myself, but if you do a web search for "picasa face recognition", you'll find lots of information on how to use it.
The program I use for storing all my photos is Windows Photo Gallery. It uses faces I've marked as a particular person, and suggests faces in other unmarked photos that may be the same person that I can also mark. I think it works quite well and I find it helpful for identifying relatives in pictures. It does indeed find similar faces who may or may not be the same person. Of course, no system will be perfect. You will want to inspect and decide for yourself about any automated selection.
The genealogical company that first introduced face detection/recognition to genealogists was MyHeritage using the technology originally developed by Riya about 2006. You can make use of the technology from the family site that you create with their free Family Tree Builder software. As they say in (an older version) of the User Guide:
"Photos uploaded to your family site can utilize our face recognition technology for genealogy. By sharing millions of facial templates among users, it helps you find photos of your ancestors in other users' photo albums, identify mystery people in your photos and find relatives based on facial similarities."
If you and your cousin want to go a bit further and put those pictures up on MyHeritage's system, you might match their pictures to those of other researchers.
MyHeritage even has a (somewhat cheeky) Celebrity Look-Alike's app that uses the same technology.
And if you're going to do all this, you might as well also try out Microsoft's How-Old.net website.
FamilySearch has introduced a tool to see which ancestors you look the most like:
You upload a photo of yourself and then your ancestors.
As a long-time user of Picasa and Google Photos, I have to say there's a new approach that's pretty cool, especially as a way to sort through family photos. It's also designed specifically for sharing family history.
Ponga.com uses computer vision to sort through piles of digital images so that you can quickly organize them by the people in the picture. It's especially useful for sharing family history because it gets you to the good part of telling stories in pictures quickly.
Your original images can be arbitrarily large (.tif, .png, or .jpg) and protected with archival storage. Working copies are used in creating Ponga pictures so your original files with metadata and file names are never modified. Pictures themselves become links where you can associate all kinds of written stories, media and associated images to pixel ranges on the image.
Disclosure: Before I catch myself going on about features, I should disclose the obvious that I'm part of the founding team.