DcnnGrasp: towards accurate grasp pattern recognition with adaptive regularizer learning

Xiaoqin Zhang, Ziwei Huang, Jingjing Zheng, Shuo Wang, Xianta Jiang

Abstract: To address the challenge of handling unseen objects in grasp classification, we propose a novel dual-branch CNN (DcnnGrasp) and a new training strategy based on JCEAR. The simulated experimental results demonstrate the significantly superior performance of the proposed method to other state-of-the-art methods in grasp classification. In addition to its superior generalizability to unseen objects, our method also exhibits stronger robustness in handling 3D information of objects and datasets with only a few grasp labeling results. Therefore, introducing the object category information by combining dual-branch networks and the JCEAR training strategy holds great significance and application prospects in the field of grasp pattern recognition.