Learned Uncertainty Calibration

A collection of Python scripts that connect to SLAM systems and analyze their covariance estimates on various datasets.

Currently supports XIVO with the following datasets:

• TUMVI VIO
• XIVO dataset distributed with XIVO

Setup

1. Install dependencies: numpy, scipy, matplotlib.
2. Build XIVO. Ensure that cmake uses the same version of Python 3.
3. Set the environment variable XIVO_ROOT to the root of XIVO source code.
4. Follow instructions in this file

Replicating the paper

Commands for executing the experiments in ``Learned Uncertainty Calibration for Visual Inertial Localization" are here.