There is a nice review of the literature in the introduction to this journal article (on Australian students), including research, largely from the '70s, that found teachers to be about 10-50% accurate (depending on the way accuracy was measured) at ID'ing GT children, while parents were about 50-96% accurate.

Note that the research is mostly on preschool and kindergarten-age students.

http://files.eric.ed.gov/fulltext/EJ750772.pdf

This is the key article generally cited:

https://www.researchgate.net/public...d_children_as_a_function_of_school_level

...pronounced like the long vowel and first letter of the alphabet...