Aiming at accuracy of robotic arm grasping pose detection, a novel generative grasp detection network LGAR-Net (loss-guided attention mechanism and residual network) consisting of deep semantic extraction module and feature recovery module is proposed. The deep semantic extraction module combines the residual block with the loss-guided attention block to improve the ability of the network to select the best grasp pose in the process of grasping irregular object. Based on the RGB map as input, the experimental findings demonstrate that 97.7% accuracy can be obtained on the Cornell public dataset. On the Jacquard dataset, an accuracy of 94.7% is attained in the meantime.
Grasping Pose Detection Based on Loss-Guided Attention Mechanism and Residual Network
2023-10-11
2716126 byte
Conference paper
Electronic Resource
English
Robust grasping under object pose uncertainty
British Library Online Contents | 2011
|Stable grasping under pose uncertainty using tactile feedback
British Library Online Contents | 2014
|Stereo-Vision-Guided Object Grasping
British Library Conference Proceedings | 1999
|NTRS | 1987
|