12 datasets found

The dataset used for the experiments in the paper, consisting of 2D and 3D environments, and objects to interact with.

The Gibson dataset is used to train DRL agents for visual navigation. The dataset consists of RGB images and depth maps of indoor environments.

The RoboTHOR dataset is a simulation environment for embodied AI research, containing 15 validation environments with 12 different goal object categories.

The ALFRED benchmark includes 25,743 trajectory-instruction pairs, covering 7 different task types with varying levels of complexity.

The Goal Drift Dataset is a collection of triplets (current observation, goal imagination, instruction) from the OpenAI Contractor Gameplay Dataset, used to train the Imaginator.

Homerobot: Open-vocabulary mobile manipulation.

ObjectNav revisited: On evaluation of embodied agents navigating to objects.

Object manipulation via visual target localization.

A benchmark dataset for MultiON task across different houses for evaluation and future use by researchers.

The dataset used in the paper is the AI2-THOR dataset, which is a 3D indoor environment dataset. The agent must navigate from a random starting position to a pre-specified...

The Habitat dataset is a large-scale indoor simulator dataset containing 145 semantically-annotated indoor scenes.

The HM3D dataset is a large-scale 3D environment dataset, consisting of 1000 large-scale 3D environments for embodied AI. The dataset is used to evaluate the performance of SLAM...