IR ICRA 2014 – Illumination-Robust Visual-Inertial Dataset
The IR ICRA 2014 dataset provides visual–inertial recordings designed to evaluate illumination-robust monocular visual odometry and SLAM algorithms. It contains challenging sequences with significant lighting variations, fast exposure changes, and difficult low-light conditions that commonly occur during indoor and outdoor flights of micro aerial vehicles (MAVs).
These sequences are useful for benchmarking:
- illumination-robust feature tracking
- monocular/multimodal VO under high dynamic-range conditions
- visual–inertial fusion methods
- exposure-robust perception and real-time onboard estimation
Data Access
➡️ ETH Research Collection (landing page & downloads): https://doi.org/10.3929/ethz-b-000721641
Contents
The dataset includes:
- Monocular camera recordings under strongly varying illumination
- Inertial measurements (IMU) aligned with camera timestamps
- Sensor calibration (intrinsics/extrinsics)
- Sequences with rapid brightness changes, shadows, and dark transitions
- Ground-truth alignment or reference trajectories for evaluation (where available)
Reference Paper
Illumination-Robust Monocular Visual Odometry for On-Board MAVs Presented at the IEEE International Conference on Robotics and Automation (ICRA), 2014. Authors include contributors from the Autonomous Systems Lab, ETH Zürich.
Tags: visual-inertial, illumination, calibration, slam