模型222222222

# DEX-NET 2.0 GRASP QUALITY CONVOLUTIONAL NEURAL NETWORK (GQ-CNN) WEIGHTS # OVERVIEW This directory contains the weights of the GQ-CNN trained on 6.7 million point clouds, grasps, and grasp metrics from Dex-Net 2.0 generated from 1,500 3D object models from the KIT and 3DNet datasets. The network was trained on January 24, 2017. For more information on the dataset, architecture, and training method, please see our paper at https://arxiv.org/abs/1703.09312. The model is referred to as "GQ" in the Experiments section of the paper (Section VI). If you use the weights of this network in a publication, please cite: Jeffrey Mahler, Jacky Liang, Sherdil Niyaz, Michael Laskey, Richard Doan, Xinyu Liu, Juan Aparicio Ojea, and Ken Goldberg. "Dex-Net 2.0: Deep Learning to Plan Robust Grasps with Synthetic Point Clouds and Analytic Grasp Metrics." Robotics, Science, and Systems, 2017. Cambridge, MA. # USAGE This model can be used with the gqcnn python package developed by the AUTOLAB at UC Berkeley. Please see https://github.com/BerkeleyAutomation/gqcnn