The Aria Synthetic Environments Dataset has been procedurally generated to produce a diverse set of interior scenes. Each scene has a unique room graph connecting multiple rooms, and unique placement of architectural features, such as windows, doors, and pillars.

Each of the 100,000 unique scenes, is filled with objects from a digital library, each with high-quality materials and geometry. Objects are diverse and placed according to a simple set of rules that result in a physically-valid location for each object.

Each simulated sequence is rendered to reflect the sensor characteristics of Project Aria glasses, including simulated lens and sensor characteristics. Inertial data is also simulated using a noise model that reflects Project Aria’s IMU sensors.

Architectural features, such as doors, windows, and pillars, are described with a CAD-like language, including the feature type, location, and dimensions. This unlocks new exciting ways to tackle research challenges related to reconstruction and detection tasks.

Before rendering each sequence, device trajectories are simulated within each environment according to a set of rules that mirror how users walk while wearing Project Aria glasses. Trajectories are created automatically and ensure a full traversal of each virtual scene.

In addition to per-frame depth and instance maps for each sequence, semi-dense point cloud representations are also made available for each environment. These additional representations open up new ways for researchers to tackle reconstruction and detection tasks.