I was reading through a paper, and came to the section where the researcher stated her criteria for selecting video segments. I thought it may be useful to collaboratively compile a list of the types of criteria that researchers are looking for. Comparing the list may be an illuminating way to identify some features of a broad but specifiable algorithm for video analysis. (So Ramsin can make the sweetest algorithm ever).
“Videos were selected based on the presence
of a target female performing an individual foraging event,
as well as on the following designated video acceptance
criteria: (a) the individuals were identifiable; (b) it was
possible to identify the beginning and end of the chase
sequence; (c) the prey was visible, or if the prey was not
completely visible, it was possible to identify the position
of the prey based on the behavior of the dolphin; (d) if the
calf was present, the calf was in a nearby position (within a
proximity of two body lengths) from which it was capable
of observing its mother; and (e) if the calf was present, it
was possible to identify the position of the calf relative to
the mother during the foraging event.”
From Bender, C.E., Herzing, D.L., and D.F. Bjorklund (2008). Evidence of teaching in atlantic spotted dolphins (Stenella frontalis) by mother dolphins foraging in the presence of their calves. Animal Cognition.