6 replies to this topic

[mcarey]

I hope that this is not off of the subject of scientific photography.
I am looking for some advice and help finding some people/programmers to help build a computer program to use digital photos to do facial identification of sea turtles.
Again I hope that I am not out of line with this post.

[Giles]

http://www.myheritag...Recognition.php


they have face recognition software.

takes a photo of you and gives you images of celebrities you look like.

Apart from that I personally am of very little help for you.

[mcarey]

http://www.myheritag...Recognition.php
they have face recognition software.

takes a photo of you and gives you images of celebrities you look like.

Apart from that I personally am of very little help for you.

Giles,
Thanks for the reply and the link.
I am actually looking to find more along the lines of pattern recognition.Every sea turtles face has a unique set of patterns much like a fingerprint.I am trying to find a software program that already exists or someone who can write one.
With the ease and quality of digital photos this will be a great research tool.
Again thanks for the link.
Michael

[bvanant]

I hope that this is not off of the subject of scientific photography.
I am looking for some advice and help finding some people/programmers to help build a computer program to use digital photos to do facial identification of sea turtles.
Again I hope that I am not out of line with this post.

There are a couple of ways that this is often done. In our lab we use the pattern recognition tools toolbox from matlab by Bob Duin (PRTOOLS). For some of our experimental data analysis we tried to use PITTPAT but we were looking for blood vessel ID not faces. Accusoft has a bunch of software for both facial and pattern recognition, they have tried to get us to buy some but we haven't looked at it more than cursorily.
Bill

[Glasseye Snapper]

Hi Michael,

CNN ran a story a few days ago about the use of a robotic camera to help catch an image of the illusive ivory billed wood pecker somewhere in the Sourthern USA. If memory serves me well, the camera took 20 shots per second with software evaluating each image and keeping only the ones that may have the bird on it. This seems to be a relatively high profile project but they have had little luck with many false positives in the form of falling leaves and other objects mistaken for the real thing.

These cameras still had it relatively easy with a stationary background and only one object to recognize. Doing this to sort many fish species, possibly with shots taken at a variety of locations, different distances (and thus colors) etc seems a horrendous challenge to me. It may be cheaper, faster and more accurate to pay a bunch of summer students $8 per hour to flip through images and sort them. Now that I think about it ... I am flipping through fish images all the time just for fun. Maybe I can make some money off of you

Bart

[bvanant]

These cameras still had it relatively easy with a stationary background and only one object to recognize. Doing this to sort many fish species, possibly with shots taken at a variety of locations, different distances (and thus colors) etc seems a horrendous challenge to me. It may be cheaper, faster and more accurate to pay a bunch of summer students $8 per hour to flip through images and sort them. Now that I think about it ... I am flipping through fish images all the time just for fun. Maybe I can make some money off of you

Bart

Actually turtles will be a lot easier than birds, you have a lot of leeway in how close you get and the patterns are quite distinct. In our case, we are interested in imaging of new blood vessels around an implant and if you want to make $8 per hour identifying them we might be able to hire you, although the last two students lasted only 2 or 3 days and fled in terror.
Bill

[Glasseye Snapper]

Ahhh turtles, should have read the original post more carefully. That will definately be more doable.

Good luck with the project!

Bart