DexNet 2.0
收藏arXiv2025-09-30 收录
下载链接：
https://berkeleyautomation.github.io/gqcnn/benchmarks/benchmarks.html
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是为了训练和验证抓取质量卷积神经网络（GQ-CNN）而设计的，其中包含了各种物体和抓取方式。此外，数据集还包括基于独特物体、姿态和图像的训练/验证分割，以评估抓取模型的性能。规模上，数据集根据物体、姿态和图像进行了多分割。该数据集的任务是机器人抓取。
This dataset is designed for training and validating Grasp Quality Convolutional Neural Networks (GQ-CNN), encompassing a wide variety of objects and grasp configurations. Additionally, it includes training/validation splits based on unique objects, poses and images to evaluate the performance of grasp models. In terms of scale, the dataset adopts multiple splits categorized by objects, poses and images. The core task targeted by this dataset is robotic grasping.
提供机构：
Berkeley Automation
