Facial Identification in Two Old Photos #latvia #photographs


Gail H. Marcus
 

I previously posted a photo (reposted here) to try to figure out if the individuals in the photo were related.  I received some very helpful observations (thanks to all who replied!), but in the end, nothing was definitive.  I have since been provided with another old photo that MAY be a later photo of one of the individuals in the first photo.  (I am not certain.)  I was hoping that someone with good facial recognition skills might be able to see if one of the individuals in the later photo resembles anyone in the earlier photo.

The individual I want to try to match is the man on the lower right side of the photo of 5 people.  The question is, do his facial features resemble any of the men in the photo of 3 couples, and if so, which man?  (Note that the photo of the couples was probably taken between 1905 and 1920, and the photo of the 5 people was probably taken in the late 1930s.) 

Thanks for any help anyone can give me.



Gail Marcus
Bethesda, MD


lesleyedwards@...
 

Looking at the clothing and hairstyles etc in your earlier image, the photo is definitely before 1914 and much more likely to be c. 1905-1910.

Lesley Edwards
Cheshire, England


Odeda Zlotnick
 

I'd say he could be the man on the right in the upper picture - more so than any of the others.
Something about his eyebrows, and the angle and size of his ear (relative to his face) and the slight tilt of his head to the viewer left (his own right) make think so.
--
Odeda Zlotnick
Jerusalem, Israel.


rroth@...
 

Hard to say given the passage of twenty years and the fact that two of the men in the older photo are well-bearded but here goes. The one in the older photo whose chin we can see, the seated one, does seem to have the same chin and eyebrows as the oldest man in the newer picture. The shorter standing man seems to have a chin that is too pointy to be a match, but the taller one looks to have the same chin as the subject.

Side note: comparisons only work within a single picture. Someone else here mentioned the "upper" picture; on my screen they are side-by-side.

==========
Robert Roth
Kingston, NY
rroth@...


Gail H. Marcus
 

Thanks for the responses so far.  Very thoughtful.  I also appreciated the help dating the photo.  That helps a lot.

FWIW, I noticed that the orientation of the photos on my screen changes depending on how wide you make the window--so when it was full screen, the photos were side by side, but when I narrowed the window to half the screen width, the photos shifted and were stacked, with 5-person photo on top.  So maybe best to refer to the 6-person photo and the 5-person photo. 

Finally, just to add to the confusion, I had one private response, so I won't include any identification, but it says "I would say that in the middle..."  So any further analysis might be helpful.

Thanks again, and happy Thanksgiving to those who are celebrating today.

Gail Marcus
Bethesda, MD


Lowell Nigoff
 

Has anyone used the Google Photos to compare faces. 
Lowell Nigoff
Lexington, KY 


Shelley Mitchell
 

On the top picture, it appears that each man is with his wife. The wife that matches the most closely to the wife on the bottom is the one on the left. Maybe that can break any ties. 


Shelley Mitchell, NYC


angel kosfiszer
 

The man on the right on the older photo was loosing his hair and is the bold man on the new photo. His ears match on both photos. His wife looks alike. I would say the new photo is of the couple on the right of the old photo, with their children.
--

Angel Kosfiszer

Richardson, Texas


Jerry Scherer
 

Try using  FamilySearch Discovery - Compare-a-Face — FamilySearch.org. See examples below.











Jerry Scherer
Toronto, Ontario


Gail H. Marcus
 

I haven't used either the FamilySearch Compare-a-Face software that Jerry used or the Google Photos software that Lowell mentioned, so Jerry's comparisons are very interesting.  BUT...to me, the 20-25% matches seem low.  That would seem to me to apply more for siblings than for the same person.  So my question is--how good is the software for matches of the same person 20 or so years apart?  For a "real" match (for example, if I ran 2 pictures of what I was sure was the same person), should I expect numbers higher than 20-25%? 

Gail Marcus
Bethesda, MD


Esther Brill
 

The Jewish Genealogical Society of Cleveland is having a guest speaker related to this topic.  Perhaps you can register or find out how to contact the speaker lots of good luck

 

Using Facial Recognition Tools to Identify Unnamed Ancestors for Genealogical Research"
Presented by Scott Genzer

Registration Process

This program is free and open to the public, but space is limited.

Priority will be given to members of the Jewish Genealogy Society of Cleveland (JGSC).

Preregistration is required and must be requested by 12:00 Noon (EDT) on December 1st.

Registration Process

This program is free and open to the public, but space is limited.

Priority will be given to members of the Jewish Genealogy Society of Cleveland (JGSC).

Preregistration is required and must be requested by 12:00 Noon (EDT) on December 1st..

  • JGSC Non-Members: Please click on the following link: rsvp@... and
                                          provide your full name, email address, and complete mailing address. 

Look for an email acknowledging receipt of your request. Then on Tuesday afternoon, the day before the program, the Zoom link will be forwarded to you, along with the meeting number and passcode.

Esther Levine Brill



 

 

 

 


Eva Lawrence
 

Gail Marcus writes that the 25% certainty for deciding two photographs
are the same person seems to her to be low, and asks about the
reliability of the comparison software.
I'm replying as a mathematician and former computer programmer, but
without any knowledge of this particular site.
To evaluate the percentage of certainty, she needs to think about the
amount of information she has given to the software in order for it to
make the judgement and what the number is telling her.

To obtain a numerical result, the site needs to use numerical
measurements. The dimensions of the subject's features in each image
determine the the number such software finally comes up with.
Digital photographs lose resolution every time they are digitally
processed
The resolution of the images determines the accuracy of measurements
made.
So maybe the software is numerically honest about the reliability of
its answer given the digital evidence presented.
Comparison software is no more accurate than a human expert can be, just
quicker and more easily available,.
She could try the comparison again making sure she presents images of
the highest possible quality.

After all, the software must have been designed by just such experts.
Eva Lawrence
St Albans, UK
--
Eva Lawrence
St Albans, UK.


Jerry Scherer
 

I decided to test your question.
 I compared my mother's family picture taken in 1933 with her picture taken in 1948, 1978 and 2003 (her last picture - she died 6 months later). I then compared her, her mother (my grandmother) and 2 sisters (from the same 1933 family photo) to determine how well they resemble each other.   I had great success using FamilySearch software to test family relationships. I will leave you to decide from the examples below. This tool is a fun exercise as it generates probabilities and should not be used as proof of the relationship.

1. Mom (16 yrs) in 1933 and 1948



2. Mom in 1933 and 1978


3. Mom in 1933 and 2003.


4. Mom (16) and sister (19)


5. Mom (16) and sister (25)



6. grandmother (55) and mom (16)



7. grandmother (55) and mom (31)


8. grandmother (55) and mom (61)




9. grandmother (55) and mom (86)



Jerry Scherer
Toronto, Ontario


Gail H. Marcus
 

These comparisons using other people over time are very helpful and give me just the data points I need to help me make sense of my photos.  From this, it looks like the software does a good job of matching people, even as they age, so my initial assumption about who is who in the picture of 6 people is probably wrong.  Disappointing, but it's important to know.  I now have to hope I can find more photos and see if I can figure out who they are.

Gail Marcus
Bethesda, MD


Odeda Zlotnick
 

Very impressive!
Do you, Gail, or you Jerry have pictures in which you can compare a male whom you know, in pictures with and without a beard?  It would be interesting to see how the software handles that.
--
Odeda Zlotnick
Jerusalem, Israel.


Susan stone
 

I KEEP vacillating.  So I would try a facial recognition software like photomyne.  If you write back I can do it for you as I have a subscription.
susan stone
Evanston, IL


jbonline1111@...
 

I photographed a photo of my son at 16 today to send to him.  I also uploaded it to Google Photos, and it automatically matched the early photo to one of him this year at age 50. I would definitely upload these photos to Google Photos and see what happens.  
--
Barbara Sloan
Conway, SC


Jerry Scherer
 

I do not have any pictures in my collection that compare two males, with and without a beard. My guess is that it will not make a difference. My assumption is that the software compares many facial "landmarks" such as distance between eyes, nose shape and size, jaw structure, face shape (round or oval), etc. Hair would not be a factor into the algorithm. In my mother's case, her hair became thinner as she aged and the software still managed to correctly identify her after 70 years. If you have two or more photos to compare, I'd be happy to test this hypothesis. I created a primitive facial software application about 10 years ago using this principle. As I expected, professional software developers have come up with much better algorithms. You may want to sign up to Scott Genzer presentation at JGS Cleveland, posted by Esther Brill above. Scott presented at our society, JGS Toronto, on Feb 28, 2021. His presentation was excellent.

Jerry Scherer
jscherer@...
Toronto, Ontario


Gail H. Marcus
 

Thanks to Odeda Zlotnick for her suggestion to try to match a bearded and clean-shaven version of someone I know, and to Jerry Scherer for his help in comparing the pictures I sent him, I think we can conclude that the software does, indeed, ignore facial hair, and instead, looks for other facial features, and that it can identify a person regardless of beards and mustaches.  For personal reasons, I don't want to post the photos.  However, for everyone's information, I can report that I sent Jerry "then and now" photos taken about 43 years apart.  There were 4 men in the earlier photo, so he was able to test the match between the bearded man in the more recent photo and each of the clean-shaven men in the older photo.  I felt that would allow us to see if there was a distinct difference between the right pairing and the wrong one.  The software gave a 91% rating to the match that I know is right, and ratings of only 17-21% when comparing the bearded man to the other people in the photo.

So I think this confirms that the software is reliable.  Which means that the man I was asking about in one of the photos I posted earlier is not a match to anyone in the other photo (all the percentages were low in that case).  This is a personal disappointment, as I was hoping that the photo with the bearded men would prove to be my family, but it is more important that I have the right answer than to continue making a wrong assumption.  I'm very grateful to Jerry for his help, and to everyone for their thoughtful suggestions.  

Gail Marcus
Bethesda, MD


Deborah Wiener
 

I have two photos-one I know to be my grandfather taken during the war, and one from 1904 that I think could be him. Is anyone able to assist in a comparison?

Debbie Wiener

dwiener@...

Melbourne Australia