Evulation framework for name calling response

[lucie271]

Description

Analysing the automatic annotation for Reponse to name calling from the videos, requires:

• audio processing (and speech analysis)

• video processing (gaze estimation)

• evaluation framework to check the performances

• Build framework for evaluation :

• split SAILS videos in train/test for

1. name calling detection
   Check performances of 'person entity" extraction from speech analysis. Was there someone calling the child in the video ?
2. reaction when name calling identified in video
   Check performances of timepoint matching between name calling and sudden gaze changes. Does the child change where he's looking when being interepelled ? Is this analysis good enough for Response To Name analysis ?