7 datasets found

The Talk2Car dataset is a benchmark for language grounding for autonomous vehicles. It contains images of urban scenes and natural language commands referring to objects in the...

The dataset used in the paper is a collection of natural human instructions for a 3D room containing everyday objects. The instructions are used to train an agent to follow...

The dataset used for training the InstructVid2Vid model, which consists of video-instruction-edited video triplets.

The REVERIE dataset is a dataset of household tasks in an indoor environment. It contains images annotated with natural language instructions including the referring expressions...

The dataset used in the paper is a set of 35 complex and ambiguous object goal navigation and mobile pick-and-place instructions.

A novel benchmark for text rewriting with a focus on cross-sentence rewrites, covering a wide variety of rewriting types expressed through natural language instructions.

The dataset used in the paper is a set of 55 training instructions and 15 test instructions for zero-shot evaluation.