Posted: 14 Aug 2014 | Author: Sue Adams | Filed under: Genealogy issues, Photo analysis, Research strategy, Sue's family research | Tags: Allport, discrepancies, Face comparison, face recognition, family history, genealogy, identity, Photograph Identity, PIQ, poll result, University of Strathclyde, Uttoxeter |
Thank you to the 21 people who responded to Photograph Identity Questions 4, 5 & 6. A couple of people did not answer all three questions, so each ended up with 20 answers as follows:

Face comparison 4,5 & 6 results
The no votes have it, to varying degrees, with 70% for photo pair 4, 60% for photo pair 5, and 90% for photo pair 6. From these results, pair 6 certainly seems to be different people, but it is hard to be so sure for pair 4 and pair 5.
In this series, I have now asked ‘Is this the same person?’ for 6 pairs of faces. Photo pair 1 was a control case where I knew the answer was ‘No’. Photo pairs 2 to 6 are all about comparing people in the group photo below with other photos in the same album.

Allport, Thomas B. (photographer, Uttoxeter). ca. 1860s-early 1870s. Group of 8 people. Faces extracted for comparison in photo pairs 2-6.
To recap all 6 results:
Photo pair |
Result |
Result% |
Strathclyde result |
1 |
No |
76% |
Human: Yes 69%; Picasa: Yes, threshold 65 |
2 |
Yes |
94% |
Human: Yes 94%; Picasa: Yes, threshold 75 |
3 |
No |
65% |
|
4 |
No |
70% |
|
5 |
No |
60% |
Human: Yes 82%; Picasa: Yes, threshold 65 |
6 |
No |
90% |
|
Three of the photo pairs were included in a project I undertook in 2010 for the Genealogical Studies postgraduate program at the University of Strathclyde. The methods were a little different as most respondents gave their answers offline and they were asked to give ‘instant’ answers rather than try to consciously analyse the photo pairs. In the Strathclyde study, 4 of the 5 control face pairs known to be the same person scored 70% or more ‘Yes’ votes, but only 1 of 4 the control face pairs known to be different people scored over 70% of ‘No’ votes. So I thought responses to face pairs of unknown identity with a substantial majority of ‘Yes’ votes, especially those over 80%, were likely the same people. I hoped for consistent results for the 3 pairs repeated in this series. Photo pair 2 delivered the same result, but photo pairs 1 and 5 did not.
Comments from respondents suggest that they spent time consciously analysing the photo pairs. If undecided did you vote ‘No’? Did you become less sure the longer you tried to analyse the photos? Are these potential reasons for the preponderance of ‘No’ votes? Could the 60-70% middle ground ‘No’ votes really indicate uncertainty? I welcome comments on these questions.
I admit that I hoped that pairs 2 – 6 would have decisive ‘Yes’ answers to support the conclusions summarised in Cartes de Visite album links to the Stanley family and Earls of Derby.
© Sue Adams 2014
Like this:
Like Loading...
Posted: 20 Aug 2013 | Author: Sue Adams | Filed under: Genealogy software and data, Photo software, Sue's family research | Tags: 50 Marriage Mondays, Adams, Birmingham, Canning, face recognition, family history, identity, metadata, Ward End |
When I first got my hands on this photograph it had no helpful annotation on the back to tell me who the people were. The photo is part of my grandmother’s collection, so I could recognise her and my grandfather. As they died many years ago, I could not ask them to identify the other people. Three people of the next generation, the baby in his mother’s arms, the toddler on her grandmother’s knee and the child bridesmaid are still alive and all recognise the groom as a favourite: Uncle Albert.

The marriage certificate provides the date, location and names three people present at the wedding in addition to the couple:
Bride: Margaret Canning
Groom: Albert Adams
Date: 12 April 1941
Location: St Margaret’s Church, Ward End, Birmingham
Father of Bride: Arthur Solomon Canning
Father of Groom: Thomas Henry Adams (deceased)
Witnesses: A S Canning, J Adams
Although some of the people in this picture have been identified, others have not. First, I want to do the electronic equivalent of writing on the back, and second, I want to share the photo with relatives and let them add to the annotation.
Annotating Digital Images – Metadata
It is important that my annotations are embedded in the image file and that they are not lost when the file is copied or edited. Two commonly used file types that support embedded text are jpeg and tiff, which also support information like camera settings, date and time of creation, and copyright. Rather than describe the positions of people, which can get cumbersome, I want to point to a face and label it with the person’s name.
Digital cameras, social media and image processing software now commonly boast ‘face recognition’ capabilities. Two processes are often conflated when people talk about face recognition. The first is the ability to determine that a face is part of the picture, rather than some other object with similar dimensions (e.g. a ball, balloon). This problem has been solved and successfully implemented in many cameras and software, which identify the part of the picture containing a face and highlight the region in a rectangle.
The second problem, the ability to compare two faces and determine whether the same person is depicted, is much more complex and difficult. Automatic comparison and identification requires multiple images to train the software to recognise a person. The training is done by a person. People are talented at recognising other people, computers aren’t.
All the embedded information, the file’s metadata, is the needed for the person labelling functions I want to work.
Picasa and Photoshop Elements – Metadata compatibility
Two image processing programs, Picasa 3.9 (free) and Photoshop Elements 8 (came bundled with some hardware) installed on my computer, are both capable of identifying face regions and labelling them. However, faces labelled in one program are not recognised by the other.

Screen shots of face tagging in Picasa (top) and Photoshop Elements (bottom)
There are many ways labels, tags and definition of face regions can be implemented by software, so programs have developed a variety of different solutions. Incompatibility between programs is a consequence. Consumer dissatisfaction prompted a consortium of digital media companies, The Metadata Working Group, to publish technical guidelines in November 2010, aimed at overcoming the incompatibilities.
Photoshop Elements 8, released in 2009, does not seem to store face regions in the image file. Photoshop Elements is now on version 11, so it might have implemented the metadata guidelines. Picasa 3.9, the current version, does store face regions in the file metadata, but they are not recognised by Photoshop 8.
Face regions are stored separately from tags. Tags are widely used to facilitate searching files containing tag labels. For example, photos depicting Albert tagged as ‘Albert Adams’, can be found from my operating system or image software. Photoshop created tags as I labelled face regions, but Picasa did not. It turns out I want both.
For now, I prefer using Picasa for naming people as it is more user-friendly, but use Photoshop for other image editing tasks.
Sharing and online collaboration
I would like to share this photo online in a way that allows fellow genealogists or relatives to tag the unidentified people.
Social media sites such as Facebook and Google+ have face region and tagging capabilities. However, only people with whom you are associated on the website can be tagged for reasons of online privacy and social etiquette. Most of the people in this photo are long dead and certainly not on social media!
Picasa has a facility to upload photos, which is in transition from the old ‘Picasa Web Albums’ to ‘Google + Photos’. I uploaded the photo and viewed it online, but am not sure which service was in operation when I could see this in my browser:

So, dear relatives, can you identify any of the people not yet tagged?
© Sue Adams 2013
Like this:
Like Loading...
Posted: 02 Sep 2010 | Author: Sue Adams | Filed under: Analysis, Photo dating, Sue's family research | Tags: face recognition, Fleming, photograph collection, Stanley |
The Carte de Visite album passed down from Mary Fleming (nee Lamb) has been a wonderful example of an early photograph collection to work on. This collection dates from the late 1850’s to the 1870’s, spanning the period of great popularity of the format for collections of celebrity pictures, exchanging with friends and acquaintances in place of the visiting card and family keepsakes.
I have established the connection between the Fleming and Stanley families. In the 1868 London Post Office Directory’s Court section, Charles James Fox Stanley is listed at 9 & 10 Half Moon Street, the address of Flemings Hotel.

The four photos annotated with names Stanleys: Evelyn Emma, Margaret Alice, Douglas James George and Albert Hamilton are all consistent with being those members of the Stanley family, children of Charles James Fox Stanley, 3rd son of the 13th Earl of Derby. I dated the photos based on costume, props and records of the photographer’s activities found in census, newspapers and trade directories, estimated the ages of the subjects and calculated which Stanley family members could be represented in the album. The Stanley family were resident at Loxley Hall, near Uttoxeter in 1871. The group photo portraying eight people was taken by T B Allport of Uttoxeter.
Then I compared the faces, eliminating bias by presenting pairs of faces to a sample of genealogists, friends and family, and using Picasa to compare faces. As I expected people performed better than the computer software. These techniques provided supporting evidence for the identity of some Stanley family members, but sadly disproved my hopes of finding one of the Earls of Derby in the album.
Like this:
Like Loading...