[Retracted] A Secure and Efficient Multi-Object Grasping Detection Approach for Robotic Arms

Figure 2

The figure shows the general technical flowchart of our approach. The input image is collected at the edge, compressed by the encoder, and then uploaded to the cloud. The image will be reconstructed by the decoder in the cloud. Then grasp parameters are obtained through classification and regression networks to get the bounding box. According to the probability, the most likely bounding box is selected as the final result of grasp detection. The blue box and green box in the figure represent the edge end and cloud end, respectively.