TartanAir Dataset Documentation

Install:

pip install tartanair

Welcome to TartanAir V2!

Let’s go on an adventure to beautiful mountains, to dark caves, to stylish homes, to the Moon 🚀, and to other exciting places. And there is more! You, your models, and your robots can experience these worlds via a variety of sensors: LiDAR, IMU, optical cameras with any lense configuration you want (we provide customizable fisheye, pinhole, and equirectangular camera models), depth cameras, segmentation “cameras”, and event cameras.

All the environments have recorded trajectories that were designed to be challenging and realistic. Can we improve the state of the art in SLAM, navigation, and robotics?

🆕 TartanGround: A Large-Scale Dataset for Ground Robot Perception and Navigation

We are still working on adding the TartanGround dataset to the documentation. The tartanair python package will support this new dataset as well.

Getting Started:

• Installation
  • Requirements
  • Known Installation Issues and Solutions
• Examples
  • Download Example
  • Customization Example
  • Dataloader Example
  • Data Iteration Example
  • Evaluation Example
  • Flow Sampling Example
• API Reference
  • Initialization
  • Download
  • List
  • DataLoader
  • Customization
  • Visualization
  • Trajectories in NumPy
  • Evaluation
  • Data Iteration
• Modalities
  • Raw RGB image
  • Raw depth image
  • Raw semantic segmentation
  • Raw camera pose
  • LiDAR
  • Fisheye and Panorama
  • Optical flow
  • IMU and noise model
  • Event Camera
  • Occupancy Map
• Environments
• Troubleshooting

License

Similar to TartanAir V1, the TartanAir V2 dataset is licensed under a Creative Commons Attribution 4.0 International License and the toolkit is licensed under a BSD-3-Clause License.