I think the task of automatic identification of birds is far harder than the
examples of number plate recognition and eye scanning.
In those cases the subjects are all arranged facing the same way, and are of
similar size. They'd have a poor success rate if the subjects were all
different sizes, shapes and colours, facing any old direction, hiding in the
undergrowth, etc. Compare the problem to what happens to a camera's face
recognition focusing when the subjects turn sideways.
Rather than identifying species I'd be pretty happy for now if my camera could
at least work out which part of the scene was a bird, and focus on it.
I assume the subject is tongue in cheek, but it's interesting that the incident
that inspired it was one where his "fuddy duddy" id skills gave him the wrong
answer (identifying a raptor from of photo).
He complained that in the field the bird would be moving, giving him more
clues. Perhaps that's the real lesson from his article. If one feels one needs
to see the bird's flight style to identify it, why try from a photo?
Peter Shute
--------------------------
Sent using BlackBerry
|
