Add qualitative object information to detected objects #16
Description
Related to b-it-bots/mas_perception_msgs#8
In the scene detection actions (object and plane detection actions), we should find a way to fill the object colour and shape fields. We could use colour/shape classifiers here. For shape classification, I assume using the point cloud makes more sense, but I'm open for other proposals (this is a survey of shape classifiers for example).