The client is creating an advanced machine vision system capable of understanding and responding to human movement. In order to do so, they need set a 3D model “skeleton” within video frames with 3D representations of one or more persons. This is used to teach computers how to detect and understand the movement of human joints via machine learning.
A 30-person iMerit team was built in order to securely tag joints throughout the body using client provided software. Teams were trained to accurate annotate hand gestures, specific full-body movements, and facial markers.
With a 98% accuracy rate, iMerit teams processed over 10,000 diverse images per month for a year. Teams worked with still images of hands, faces, and full-body figures, and with videos of human movement.