Following the release of the Grasp-Anything dataset, we publish Grasp-Anything++, the linguistics-supported version of Grasp-Anything. This dataset aims to facilitate research of language-driven grasp detection.
You should obtain the previous Grasp-Anything dataset via this link.
The dataset is structured as follows:
grasp_instructions: contains .pkl grasp instructions of each scene. Each scene description is correlated to some… See the full description on the dataset page:
https://huggingface.co/datasets/airvlab/Grasp-Anything-pp.
