Hi Richard,
Thank you for your question.
This is indeed a scenario that happens in real life. Contrary to common objects/scenes, multiple anomaly patterns can overlap in a farmland image. A straightforward example is a weed cluster region covered by cloud shadow. Since both are patterns important to the agriculture-vision studies, we intend to retain both annotations, which result in label overlaps.
As mentioned in our evaluation metric page, we will consider a pixel prediction as correct if it matches any of the ground truth labels.
The Agriculture-Vision Team