giaIndoorLoc: Auto-Labeled WLAN and IMU Dataset for Indoor Localization

An auto-labeled dataset for fingerprinting-based indoor localization, generated via the VI-SLAM2tag method described in a 2022 IPIN conference paper. It was created by Marius Laska at RWTH Aachen University and includes raw and annotated data from an Android app, plus evaluation data and model weights. The dataset is designed to support research in low-effort labeled data collection for indoor positioning systems.

Use Cases

• Training WiFi fingerprinting models based on auto-labeled WLAN signal data.
• Developing sensor fusion algorithms based on combined IMU and WLAN data streams.
• Benchmarking indoor localization accuracy against provided ground-truth control point data.
• Reproducing experiments for auto-labeling techniques using the provided model weights and raw trajectory data.

Strengths

• Includes multiple data components: raw sensor data, annotated datasets, evaluation data with ground-truth from a Total Station, and pre-trained model weights.
• Dataset generation is linked to a peer-reviewed conference publication (IPIN 2022) and associated open-source repositories for implementation details.
• Focuses on a low-effort auto-labeling method (VI-SLAM2tag), which is a specific contribution mentioned in the description.

Limitations

• Column-level documentation is absent; field semantics must be inferred from the paper or code after download.
• Row count and overall data size are unknown, which may limit suitability assessment for large-scale training.
• Last update date is unknown; freshness unverified.

Provenance

Source

RWTH Aachen University

Collection Method

Recorded via a custom Android app (VI-SLAM2tag_app) and processed with post-processing tools.