3 datasets found

The Room-to-Room dataset is a photo-realistic dataset for vision-and-language navigation, where agents navigate through indoor environments based on natural language instructions.

The Room-to-Room (R2R) dataset is a benchmark for vision-and-language navigation tasks. It consists of 7,189 paths sampled from its navigation graphs, each with three...

The Vision-and-Language Navigation (VLN) task gives a global natural sentence I = {w0,..., wl} as an instruction, where wi is a word token while the l is the length of the...