Skip to main content

Digital Twin Catalog Dataset

About the data​

The Digital Twin Catalog (DTC) dataset provides static objects capturing sequences collected using Project Aria glasses and digital single-lens reflex (DSLR) cameras, combined with high quality ground truth data including device trajectories, reconstructed objects’ digital twin copies with geometry and material assets, aligned object poses. We also provide processed sensor data from our Machine Perception Services. Go to DTC Data Format to see a full list of the data we provide.

The DTC dataset contains 2000 accurately scanned and reconstructed 3D models with detailed geometry and material information (in .glb), 200 Aria sequences and 100 DSLR sequences recording static objects whose digital twin copies are within the 2000 released models. In the Aria sequences, there are 100 single-instance static objects each recorded under two different human walking trajectories, active and passive (casual). In the DSLR sequences, there are 50 single-instance static objects each recorded under two different lighting conditions.

Go to the Dataset Download page to get started with the sequence data. For more info on downloading and viewing captured models, see Object Models page.

Active Trajectory​

100 Aria sequences were recorded by taking a 360 walk around the captured objects. Each sequence will record one single object. The Aria glasses will stay within a relatively close distance to the object.

Passive (Casual) Trajectory​

100 Aria sequences were recorded by taking a casual walk around the captured objects. Each sequence will record one single object. The sequence will include trajectories such as walking towards the object, circling around the object and walking away from the object.

DSLR Sequence​

105 DSLR sequences were recorded under 2 different lighting conditions. The sequences were recorded using three DSLR cameras mounted on robot arms with three different viewing directions towards the object. We also provide 2 corresponding lighting environment maps for the DSLR sequences.

Documentation​

The DTC section of the wiki covers:

  • Getting Started
    • A quick start tutorial available to install Project Aria Tools Python package to run locally.
  • Dataset Download
    • A walkthrough of using aria_dataset_downloader to download the DTC sequence dataset.
  • Object Models
    • Everything you need to know about the released DTC object models including: a walkthrough of using dtc_object_downloader to download the DTC object models, and brief description of data content and how to visualize the models
  • Data Format
    • How DTC data is organized and stored
  • Tooling
    • Run our tools using an example that access DTC data in Python.