Finding Emma
This details an approach to try to find Emma Fillipoff using photographs
1). Stored on wikipedia. 2). Uploaded by the general public to Facehunter.co.uk.
This is based on the fact that someone may have taken a photograph of Emma inadvertently: she may be in the background in a photograph.I normally use method 1 to find a match to a police sketch to create an average image of the sketch and the matching photograph. I believe this approach provides a better image to search for a missing person where no photograph exists. I am still trying to get my new method publicised and accepted.
Note: I am a new amateur investigator and I am using free demo software Betaface, MS Azure and Facehunter.co.uk. I would like to be able to build a professional tool myself which would find the best Face ID matches on wikipedia to police sketches or photos following my methods.
All the images used are from wikipedia or photos of Emma Fillipoff. Emma’s mother has given permission for me to carry out this search (permission via a Facebook contact).
Finding lookalikes:
In the case of Emma Fillipoff a lot of photographs already exist. They can be used to search wikipedia for matching images (I use the Betaface demo to do this). Often I find a match to a single image which gives an initial excitement — but when tested against other images it stops being a match.
Below 60% matches
One example where an image matches one single photograph of Emma is this girl at a 2010 carnaval de Schaerbeek. But she matches generally using Facehunter by about 60%.
The image is from here:
But these images are not wasted. As they are near match images they can be used to search for similar images of Emma. For example it found this 2010 image — she is about a 40+% match to Emma in various images.
Another 40% face match to Emma (in V2 software), which can be used to search with — note wrong eye colour, but good teeth match.
And another 40% face match to Emma (in V2 software), which can be used to search with — note wrong eye colour, but good teeth match.
Above 60% matches
Another good lookalike is this train worker in Russia — she is a 60% match to some Emma photos:
This girl is a 68% match to one photo — but overall about a 52% match:
This girl is an 80% match to groups of photos — she is in the US army 2012.
Another reasonable match a french police cadet 2012 — she is a 60% to 90% match to some photos of Emma and a 90% match to one group of photos.
Another strong 60% to 90% match — that is found repeatedly in searches:
This girl is an 80% match to some photos:
These searches are already highlighting a pattern of finding near matches in countries like Poland, Russia, France and Germany. This suggests the name Fillipoff has European or Russian ancestry. As Emma was in Canada it might also suggest a Ukrainian background. Without knowing the family history it is difficult to assess. So I am trying to keep an open mind and keep searching very wide — not excluding a country. Emma’s mother suggests she has remained in Canada at least from 2011 and her passport was left behind in 2012. So any images of people outside Canada before/after 2012 will likely be lookalikes.
The above 90% match
This girl in France 2008 is a 100% Group match in Facehunter (using V2 software):
Another strong lookalike found this way is a high match score (100%) to a group of Emma photos.This girl is in Utah in 2009 according to wikipedia. Normally I would exclude it because of the date and location — but as the face is a such a strong match it helps develop a technique to find Emma and becomes a useful search face.
I am continuing to act as if this lookalike could be Emma: simply because of the very high score matches to a lot of the Emma pictures. I have asked Anthony Joanes (who developed FaceHunter.co.uk) to add a group of Emma images for me to check against. Note the girl has two visible lower ear piercings so an Emma fashion lookalike as well.
Facehunter shows a range of individual matches to Emma. Some are above 0.49 effectively saying the same person. Some are low scores suggesting this is not Emma. Some images of Emma herself tested this way give low scores because of poor image quality.
The image is from:
Another Utah match is this softball coach — so it could be the above 90% match is the sister of this woman — but a different eye colour:
Julie Wright at CONduit 17.png
From Wikimedia Commons, the free media repository
commons.wikimedia.org
Matching people near the lookalike
But with these high score matches normally I would have a strong suspicion that this image may be Emma or an identical lookalike. When I have this level of suspicion I test images around the candidate face.
For example a young man at the back of the picture with the Emma lookalike is a near match to a Canadian Mountie in Ottawa. A face ID score similar to a brother. The fact the image of the man found this Canadian picture is now strengthening my view that the girl potentially could have been Emma before 2012 — before she went missing — and her male friend a Canadian from Ottawa. The image on the right in the green t-shirt is from a video of a man claiming to be Emma’s boyfriend.
An average of the man in the background in Utah and the Canadian Mountie in Ottawa gives a better view of how closely their faces match:
Of course this analysis is likely to be totally wrong if Emma never left Canada.
But continuing (perhaps on a false assumption)….the girl on the left of the Emma lookalike is a also a close match to an actress Melissa Anne Smith on wikipedia.
https://commons.wikimedia.org/wiki/File:Smith_melissa.jpg
Melissa seems to have taken/has the same name as a Bundy victim in Utah in 1974.
https://www.findagrave.com/memorial/6664991/melissa-smith
Another match to the Melissa Smith lookalike was found at a Mardi Gras 2009. This simply provides another face to search with assuming it may find more so-called Melissa Smith images alongside the Emma lookalike. Assuming they are friends.
Another character in the 2009 picture matches a professional tennis player who started his career 2009. This is an average generated from two faces.
This is another average of a female tennis player who matches a girl immediately behind the Emma lookalike, that facial ID says is the same person:
Two tennis players behind our Emma lookalike in Utah in 2009 when the US open took place in September 2009. Unfortunately that probably means they would be tourists not involved in the group being photographed.
But a potential tennis circuit route 2009 for low ranked players:
Another match to a person behind the Emma lookalike in Utah. An average of an American skier with the girl in the white hoodie:
This is an average image of Emma and the Utah lookalike. The only difference you can see is the hat line.
Of course this analysis is likely based on a false premise that Emma was in Utah around 2009.
But as a first attempt at finding matches to Emma it shows my method and how in 24 hours matches can be found and the people around the potential matches further located and even given names. Finding the people in the crowd it starts to take 72 hours.
2). Using facehunter.co.uk
Note: the site owner may have taken the site down to resolve copyright issues of images from wikipedia used in another application.
Facehunter will allow the general public to test sections of their digitised photographs against a series of Emma images. This approach may provide less matches but could be used effectively to draw attention via the media. It also provides a means for me to quickly check a found face against a range of Emma pictures to help eliminate poor matches. This approach is still in development and I will add more details here as it develops.
It will allow testing of a digital photograph against individual pictures of Emma and a group of photos.
_________________________________________________________________
My method of finding dopplegangers to police sketches to use to search for missing people/criminals etc.
