Human pose detection is the foundation for many applications, particularly those using gestures as part of a natural user interface. Commercially available solutions almost exclusively use depth cameras to constrain the problem, and colour-based methods typically apply in specific use cases only (no one method can solve every set of circumstances). We introduce a novel task-based control method for human pose detection, encoding specialist knowledge in a descriptive abstraction for application by those not expert in the field, such as developers, artists and students. The abstraction holds a number of algorithms underneath which specialise either in different estimations of pose (e.g. articulated, body part) or under different conditions (e.g. occlusion, clutter). Users provide the conditions of their specific problem as a description which is subsequently mapped to the most suitable algorithm for those conditions. Using the same description, the parameters of the chosen algorithm are automatically derived. The task-based control is evaluated with images described using the abstraction; expected results are compared to actual results and show that specifying the conditions allows the system to select the appropriate algorithm to produce the required result.

Presented at the IEEE Winter Conference on Applications of Computer Vision in Steamboat Springs, March 2014.