Whiteboard documentation through foreground object detection and stroke classification

Digitization of classroom content has been receiving widespread attention in recent days. The whiteboard/blackboard contents are valuable components of classroom contents. Capturing these contents in the presence of foreground objects has been a challenging task. In this paper we propose a model for whiteboard content documentation through foreground object detection. These detected object regions are inpainted with the recent whiteboard content. This virtually provides an experience of viewing the whiteboard content without any occlusion. The resulting frames are either sent to storage as a movie file or can be streamed online to remote locations.