IDOL: 20+ Hours of Pedestrian IMU Data for Indoor Localization

The IDOL Dataset from Carnegie Mellon University provides over 20 hours of pedestrian walking data from inertial measurement units in indoor environments. Data was collected from 15 users across three different buildings using a LiDAR-Visual-Inertial SLAM system for ground truth, with synchronized IMU readings from an iPhone 8 and a Kaarta Stencil rig. All sensor readings are sampled at 100Hz, and the dataset includes structured subsets for evaluating generalization across users.

Use Cases

• Training deep learning models for orientation estimation based on 100Hz IMU time-series data.
• Benchmarking inertial odometry algorithms using high-frequency gyroscope and accelerometer readings.
• Developing sensor fusion techniques leveraging synchronized data from multiple IMUs and a SLAM ground truth system.
• Studying the generalization of localization models across different users and body types, as enabled by the provided subject IDs and user-disjoint splits.

Strengths

• Over 20 hours of recorded pedestrian motion data provides substantial volume for model training.
• Includes ground truth from a high-end LiDAR-Visual-Inertial SLAM system (Kaarta Stencil), offering a reliable reference.
• Data from 15 users of varying body types allows for studying user generalization.
• Synchronized 100Hz readings from two IMU systems (iPhone and XSens) enable comparative sensor analysis.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and total file size are unknown, which may limit suitability assessment.
• Data collection is limited to three specific indoor buildings, which may not generalize to all environments.

Provenance

Source

Carnegie Mellon University (author: Scott Sun)

Collection Method

Collected using a rigidly mounted iPhone 8 and Kaarta Stencil SLAM rig during pedestrian walking in indoor environments. Procedures were IRB-approved.

Geography

Indoor environments across three different buildings.