One-stage Robotic Grasp Detection

Robotic grasp detection is a challenging task in computer vision. Although deep learning methods have shown promising results in robotics, they still suffer from limitations such as bad runtime efficiency and restricted single-object detection capabilities. To address these challenges, we propose a simple yet effective method called Grasp-YOLOv8, which is capable of real-time multi-object robotic grasp detection. Particularly, we formulate the grasp detection problem as a standard object detection task, where grasp orientations are discretized and treated as different object classes. We evaluate our proposed method on both the widely adopted Cornell dataset and our custom CIM900 dataset. Experiment results show that our method outperforms other state-of-the-art approaches by a large margin.