DexYCB
收藏资源简介:
DexYCB数据集由都柏林圣三一学院创建,包含582,000个带有完整3D手部关节注释的2D图像帧,涉及10个对象和21个物体。数据集用于评估手部抓取预测模型对未见物体的泛化能力。创建过程中,排除了手部未接触物体的帧,确保至少有两个指尖位于物体边界框内。DexYCB数据集广泛应用于计算机视觉领域,特别是在手部与物体交互的研究中,旨在解决手势识别和物体操作的自动化问题。
The DexYCB dataset, developed by Trinity College Dublin, comprises 582,000 2D image frames with complete 3D hand joint annotations, covering 10 objects and 21 individual objects. This dataset is designed to evaluate the generalization performance of hand grasp prediction models for unseen objects. During its creation, frames where hands do not contact any objects were excluded, ensuring that at least two fingertips lie within the bounding box of an object. Widely used in the field of computer vision, particularly in research on hand-object interaction, the DexYCB dataset aims to address automation challenges in gesture recognition and object manipulation.




