Detection for Robotic Depalletizing of Interlacing Bricks Using YOLO26-OBB

This paper deals with the detection of interlacing bricks for the purpose of determining the grasp point and object rotation in a 3+1 DoF manipulation task. The proposed approach is based on the YOLO26 OBB model, which enables oriented bounding box detection and thus provides more precise information about object position and orientation than conventional methods relying on axis-aligned bounding boxes. The paper builds on earlier experiments presented in a previous study, both in terms of methodology and problem formulation. The main objective is to demonstrate and evaluate the potential of new network topologies for practical deployment in robotic grasping tasks. The experiments assess the suitability of the proposed approach for reliable localization of spacer bricks, estimation of their orientation, and determination of parameters required for grasp execution. The results indicate that the use of modern OBB architectures represents a promising direction for improving both accuracy and robustness in industrial applications where, in addition to object position, its rotation must also be taken into account.