Picasa face-recognition scan conclusions

Picasa face recognitionI have posted previously about letting Picasa 3 scan for faces so I can identify them. I had hoped to publish the results at the time but I was caught up with other things and didn’t get a chance.

Unfortunately I don’t have an accurate record of how long it took. I started it on about the 1st October with 14,000 photos to process. On the 4th it was 50% completed after I had added an additional 5000 photos because I added some of the folders under Documents. On the 5th it was saying all day that it had 51% to go. Then that evening it changed to 52%. I thought it was going to take another week, but the next day it was finished.

That’s 5-6 days. For 19,000 photos.

It ran for 24 hours a day, and I only closed it down occasionally when it was slowing down what I was doing. It used an average of 45% of my CPU, so sometimes this was a problem. I don’t remember the processor that my laptop has, but it’s a bit over 2 years old.

Of course, not all of these photos have people in them – there are landscapes, wildlife, and images of documents.

Some things I have noticed:

  • if I sign in to Google it can get the names from my contacts list
  • it runs very slowly at other times and quite quickly at others
  • it picks up faces from the covers of books and photos on the wall behind the real people
  • it can find faces in very fuzzy pictures
  • it is not bothered by hats and sunglasses
  • it quite often suggests the wrong person but that person is closely related, such as a sister, aunt or grandmother
  • it identifies people more accurately the more photos you have identified
  • it can identify people at all ages in their lives
  • it is better at identifying babies than I am
  • it doesn’t recognise cats, dogs or gorillas, although it did identify one front-on picture of a dog
  • I have a lot of duplicate photos, and when I identify one it suggests the same name for the others very quickly
  • I am terrible at remembering names
  • I nearly have more photos of my nieces than I have of my husband or myself

By the time it finished it said it still had about 6500 faces to identify. I am slowly whittling those down. I now have just over 5000. There are also the faces it can’t identify as faces, which I have to do manually if I want it done at all.

It seems to have trouble with faces if they are:

  • at an angle
  • have hair over one side
  • side-on unless they are completely from the side
  • really, really fuzzy

And yet sometimes it sees a face where there isn’t one. I thought this one must be in the background somewhere.

Panda face

He looks like he has a little beard and a receding hairline.

This is the photo it came from:

Picasa panda

Can you see the face, in the top right corner? Not a face at all!

It also picks up the hundreds of faces in the backgrounds of photos and wants to know who they are. You can mark each one as ignored, and you can see these later if you want to. When the Sydney Harbour Bridge was 75 years old they opened it to the public to walk across, and the photos from that day have many people in the background. Fortunately they are mostly wearing lime green hats so I could quickly exclude them when I saw them.

All the people in a wedding photo can be identified if you have already identified them elsewhere. Even if you don’t know their names you can give them a number, like Wedding 12, and group photos of the same person together. You can then more easily identify the person, or a relative can, when you can see a number of photos of the same person together.

I have had a wonderful time with Picasa, and I still am. I am finally learning, through having to identify photos, which of my grandmother’s three sisters is which, and what my mother’s older brothers looked like when they were young.

I have also very much enjoyed seeing pictures of the same person throughout their lives all in the one place. Here are some of my grandmother Amy Eason nee Stewart:

Amy Millicent Eason nee Stewart

You can see her from the earliest photo of her that I have, when she was a baby; as a teenager, a young mother, and so on all through her life. The photos are of varying quality but the only one I had to manually identify was the blurry side-on one in the 3rd row.

A valuable lesson I learned was in trying to identify what it is that makes this person look like that person. What is it in my face that Picasa mistakes for my grandmother’s? Or two of three nieces but not the third?

To be fair, sometimes Picasa is totally wrong. It tried to tell me that this same grandmother was in a shot of my husband posing with the Wests Tigers rugby league team. It wasn’t. When it ‘groups’ unnamed faces it tends to put faces together that are shot at the same angle. Sometimes I think it is suggesting names based on the frequency with which that name appears, or on the previously identified name, but that might just be my cynicism.

All in all I am so glad I went through this exercise. Identifying faces has become my procrastination-of-choice, and it has made me much more likely to name the faces of photos I have just taken rather than leave it for years when I can no longer remember the names. I am also determined to research the names I should know but can’t remember – school classmates, fellow safari tourists, even Wests Tigers. All those unnamed faces bother me!

52 Weeks to Better Genealogy: Week 5 – WorldCat

Week 5

Play with WorldCat.org. WorldCat is a massive network of library content that the public can search for free (user name and password not required). Not every library is a part of WorldCat, but the vast size of the network makes it an important genealogy tool. If you are looking for a specific book or publication, enter the identifying information into the WorldCat search box and see which libraries hold the item. You may even find that you can get the item through your library’s inter-library loan program. Don’t forget to search for some of your more unusual surnames and see what comes up. The goal is to play with WorldCat and examine its possibilities for your own research. If you’re already familiar with WorldCat, play with it again. The network and collection grow and change constantly. If you have a genealogy blog, write about your experiences with searching WorldCat for this exercise.

WorldCat is a catalogue of many, many libraries in the world. I’ve used it before and usually it has told me that the book I am looking for is in the State Library of NSW or the National Library of Australia. Unfortunately my genealogy society isn’t part of WorldCat, but one day that will change.

For the sake of this exercise I decided not to look for a book that I know of, but to find books that I didn’t know about. As Amy suggested, I’ve put in one of my unusual surnames – Whippy. David Whippy, born in Nantucket, Massachusetts, arrived in Fiji in about 1822 and stayed there.

So I put “Whippy” in the WorldCat search, and waited. 70 results, including a dissertation about job satisfaction in Guam University. I narrowed it down by adding ‘Fiji’, and came up with 5 results, 2 of which were the same.

The most relevant item I found was a microfilm of a play written by Isobel Whippy:

The play concerns the first British Consul in Fiji, William Thomas Pritchard, who arrived in Levuka in September 1858 and was dismissed from his post in January 1863. It is based on a theory that the Consul lost his job because of a love affair with a young woman – possibly a part-European – who gave birth to two children by Pritchard, before he married her in the British Consulate in Levuka a few days afte his dismissal. The play is in two acts – the first covering the period from September 1858 to June 1859; the second from November 1859 to July 1862. There is an epilogue concerning the year 1864.

The microfilm was published by the Pacific Manuscripts Bureau in Canberra, which I happen to know is part of the Australian National University and who microfilm manuscripts related to Pacific history. The films are available in the State Library NSW, and I have accessed them there in the past.

WorldCat, however, told me that my nearest copy was at Yale University Library, New Haven, CT 06520 United States, at a distance of 10000 miles. If I selected the other, identical title, I could find it at the State Library of NSW, the National Library of Australia, and the State Library of Victoria.

There is however, a link to Related Identities, one of which was the Australian National University Pacific Manuscripts Bureau. There’s a timeline for the Bureau that goes back to 1830, which was rather startling until I realised that most of the works listed are about American whalers in the Pacific and such, and filmed by the PMB.

So the end result of my investigation is that I can almost always find what I need in the State Library of NSW, in Sydney where I live. Anything that this library doesn’t have will probably be in Canberra and probably available on inter-library loan, although I haven’t hit this situation yet.

David Whippy didn’t arrive on a whaler but the principle is the same, so I now have a list of resources I can check to find out more about the way of life and the history of Americans in the Pacific, if not about David Whippy directly. Most, if not all, available at the State Library of NSW.

Libraries Australia has  a combined catalogue of many libraries in Australia. I don’t know if all the same libraries are in both catalogues. The free version of this catalogue is within Trove.

Trove

I put Whippy in the Search field and got a whole heap of results:

Trove - Whippy search

As you can see, there’s a vast array of stuff which will take me some time to work through. Not all of it is relevant, but some of it is. For example, the third entry under Australian newspapers (1803-1954) is a page from the Sydney Morning Herald in January 1856 containing transcripts of correspondence about American activities in Fiji. In one of the letters, written by James Calvert, the Wesleyan missionary, Mr Whippy, my David Whippy, is mentioned a number of times as arbitrating with Mr. Calvert in a dispute between the natives and an American ship’s captain. I was then able to correct the transcription of the notoriously difficult newspaper print, and download a PDF of the page or the whole newspaper.

Further down the screen there are sections for Maps, Diaries and Letters, and Archived Websites. All sections can be opened and closed on this summary screen, or clicked on to give the full list of results.

Trove is relatively new, and having now played with it I can see it is vastly superior to WorldCat for my purposes. Australian catalogues are more likely to be useful to me in general to find a book I can borrow in an Australian library. Trove gives so much more than any library catalog that I would be unlikely to go anywhere else.

It also gave me more books than WorldCat did. On its list of 96 books, journals and magazines, etc, it gives the title Gone Native in Polynesia by Ian Christopher Campbell, a book I’ve been trying to get hold of for some time. This book has a whole chapter on David Whippy in Fiji. There are tabs for each State, and under NSW I can see that it’s available at the State Library of NSW and the University of Wollongong Library. There is also a link to show where I can buy a copy – in this case from Blackwell Online for 70 pounds or Amazon from US$79.00 to US$235.00. I won’t be buying a copy for my library, but I have a search in eBay just in case.

Isobel’s play is there, with the same results – State Library of NSW, and the reference number is given.

Really, I can’t see why I would use WorldCat on a day-to-day basis. Contributers to Trove include Project Gutenberg, so I might be able to download the book I want then and there.

52 Weeks to Better Genealogy: Week 3

I’m a bit late starting on Amy’s 52 Weeks to Better Genealogy Challenge, but late is better than never, so here goes.

Week 3:

Assess yourself! You’re great at researching everyone else’s history, but how much of your own have you recorded? Do an assessment of your personal records and timeline events to ensure your own life is as well-documented as that of your ancestors. If you have a genealogy blog, write about the status of your own research and steps you may take to fill gaps and document your own life.

What do I have to document my life?

I have my birth certificate, and my marriage certificate. Those are the essentials, I guess.

I have my two university degrees, and transcripts. And my counselling diploma.

I have old journals and diaries.

I have these blogs, and their backups on my computer.

I have masses of family photos, some of which have me in them. The early ones are classified and named as best I can. They are in albums and on my computer and backed up on an external hard drive.

I tell family stories to my nieces, including my own, but I’ve realised they probably don’t really know much about me. I can change that, I guess, or I can write more of it down.

Some of this would need an IT-literate person to dig up, like the blog backups.

If I think of more, or more likely when I think of more, I’ll add them later.

Next morning

I was lying awake last night thinking about this, and I realised I was thinking of the documents and photos that a future family historian might be happy to have. I thought of some more:

  • my resume, detailing the jobs I’ve had and what I did in them
  • copies of references, from the days when written references were normal
  • a folder full of certificates of attendance and such at various courses, mostly in IT but there’s one on Thai Cooking
  • various documents and search results relating to the house we currently own
  • mortgage documents which a really keen family historian could wade through
  • a Google Map, showing where I’ve lived through my life

But the other thing I was thinking was that I could have taken this a different way. Very little of all this wonderful detail is documented in my family tree software. I use TMG, which is more than capable of handling any and all of this stuff. All I have about me personally, though, is my birth, marriage, university degrees, and attendance at various family funerals.

It has never occurred to me to try to document my life as part of the whole family history I am trying to build, and that never really part of the plan.

One good reason to do it, though, is for the practice it gives. I know more about my life than anyone else’s, and the problems I will encounter and the procedures I will have to invent will be useful when I come to document the other members of my family.

So there it is. More work to do! I knew there was a reason I was hesitant to get started on these 52 Challenges!