Palletizing or depalletizing is always based on reliably and quickly detecting objects (e.g. cartons, bags, etc.) and then passing on the determined gripping data to the robot. The basis for correct gripping is hand-eye calibration, so that the robot grips at the position where the camera sees the object. For recognition to work reliably and safely, it is necessary to detect the objects reliably.

The basis for this is the new recognition are based on Deep Learning. Once the objects are detected using Deep Learning, the grasping positions and angles are determined in the 3D image. The point cloud is then used to determine which of the objects is in the uppermost position and depalletizing begins. The neural networks predefined in the EyeVision allow this procedure for a large number of objects. If an object is not yet known, it can be quickly learned with the integrated learning tool, so that these objects are also reliably recognized.