Grasping Complex-Shaped and Thin Objects Using a Generative Grasping Convolutional Neural Network

Vision-based pose detection and grasping complex-shaped thin objects are challenging tasks. We propose an architecture that integrates the Generative Grasping Convolutional Neural Network (GG-CNN) with depth recognition to identify a suitable grasp pose. First, we construct training dataset data augmentation train GG-CNN only RGB images. Then, extract segment of tool using color segmentation method use it calculate average depth. Additionally, apply evaluate different encoder–decoder models structure Intersection Over Union (IOU). Finally, validate proposed by performing real-world pick-and-place experiments. Our framework achieves success rate over 85.6% for picking placing seen surgical tools 90% unseen tools. collected validated their pick place architectures. In future, aim expand improve accuracy GG-CNN.