Currently, only bounding boxes are supported when detecting objects. For some use cases, drawing polygons instead of square bounding boxes makes sense (for example when trying to find the total size of an object).
never-displayed