Room-across-Room (RxR) is a large-scale, multilingual dataset for Vision-and-Language Navigation (VLN) in Matterport3D environments. It contains 126k navigation instructions in English, Hindi and Telugu, and 126k navigation following demonstrations. Both annotation types include dense spatiotemporal alignments between the text and the visual perceptions of the annotators
Hey! Could somebody provide the exact size of the dataset in TB on the README file? Would be helpful in planning the location of the download on our local systems.
Hey! Could somebody provide the exact size of the dataset in TB on the README file? Would be helpful in planning the location of the download on our local systems.