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
Thank you for the visualization script. When you run the setup.py code there, you generate an args.json file that contains all the intrinsic and extrinsic matrices information during the whole time sequence. The question I would like to ask is if there is a way to get the jpg images from these extrinsic matrices. That is, given a specific time duration (i.e. from 0.5 - 6.5 sec) to get the video that corresponds to that duration (or equivalently to the sequence of extrinsic poses).
Hello there,
Thank you for the visualization script. When you run the setup.py code there, you generate an args.json file that contains all the intrinsic and extrinsic matrices information during the whole time sequence. The question I would like to ask is if there is a way to get the jpg images from these extrinsic matrices. That is, given a specific time duration (i.e. from 0.5 - 6.5 sec) to get the video that corresponds to that duration (or equivalently to the sequence of extrinsic poses).
Thank you for your time.