Understanding Human Context in 3D Scenes by Learning Spatial Affordances\n with Virtual Skeleton Models

Abstract

Robots are often required to operate in environments where humans are not\npresent, but yet require the human context information for better human-robot\ninteraction. Even when humans are present in the environment, detecting their\npresence in cluttered environments could be challenging. As a solution to this\nproblem, this paper presents the concept of spatial affordance map which learns\nhuman context by looking at geometric features of the environment. Instead of\nobserving real humans to learn human context, it uses virtual human models and\ntheir relationships with the environment to map hidden human affordances in 3D\nscenes by placing virtual skeleton models in 3D scenes with their confidence\nvalues. The spatial affordance map learning problem is formulated as a\nmulti-label classification problem that can be learned using Support Vector\nMachine (SVM) based learners. Experiments carried out in a real 3D scene\ndataset recorded promising results and proved the applicability of\naffordance-map for mapping human context.\n