Abstract: Object recognition and grabbing detection based on improved YOLOv8s is studied to improve the robot’s grasp ability of unknown objects. Firstly, the hardware platform of the intelligent grasping system is built, and the hand-eye calibration of the robot is completed. Secondly, in order to improve the detection accuracy of YOLOv8s, the CA (coordinate attention) mechanism is introduced for its backbone network. Thirdly, the mainstream lightweight object detection algorithms are analyzed and compared. The results show that the improved YOLOv8s has better detection performance, with an increase of 2.3% in mAP (mean average precision) compared to the original model, while the model size only increased by 0.1 MB. Combining the result of object detection, the depth value obtained by the camera and the grasping pose obtained by GR-CNN (generative residual convolutional neural network), the grasping classification of the parts is successfully completed. Finally, in order to realize real-time monitoring and data analysis, a host computer based on PyQt5 is developed.