SA-Grasp: A self-attention mechanism based lightweight grasp pose detection network

Read full text Summarise with VIRAT
Authors: Quancheng Pu, Hui Zhang, Lu Yang, Tieshan Li
Cited by: 0
References: 24

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In response to the current convolutional neural network-based robotic arm grasping pose detection networks being susceptible to interference from redundant information, and the tendency to misjudge object contour areas as grasping execution positions, leading to low grasping success rates of robotic arms, this paper introduces SA-Grasp, a lightweight convolutional neural network for robotic arm grasping that integrates self-attention mechanism. It reduces interference from redundant information and improves grasping accuracy. SA-Grasp achieved high detection accuracies of 98.37% on Cornell and 96.33% on Jacquard datasets, with a fast detection time of 21 ms per image. In real-world tests, it demonstrated a 94.44% success rate across 180 grasping attempts on 9 unknown objects, demonstrating its reliability. SA-Grasp is a lightweight visual grasping pose detection network characterized by high grasping success rates and fast speeds, providing a new approach for the application of self-attention mechanism in the field of robotic arm visual grasping.

Fields of Study & Subjects

Robot Manipulation and Learning Advanced Neural Network Applications Smart Agriculture and AI

Publication details

Journal / source
International Journal of Intelligent Robotics and Applications
Volume
10
Issue
2
Pages
779–794
Publication date
31 March 2026
Language
EN