The evaluation of the garbage takes place with APLASTIC-Q, a method for automatic garbage detection developed by the DFKI research unit Marine Perception, which is based on two neural networks.
APLASTIC-Q evaluates image and video material in a two-stage process. It first analyzes the image for plastic waste and then for types of waste, such as plastic bottles or polystyrene.
The two neural networks were trained using images taken for a project in Cambodia – a country for which plastic waste and its disposal is a major problem.
With Cleanup Missions by CollectiX the data-hungry neural networks is trained with new data.This enables DFKI to set up the APLASTIC-Q automatic waste recognition method in such a way that it produces more robust results with each additional application