TIE1 and SAH1: Wi-Fi RSSI Fingerprint Datasets for Indoor Localization

Two Wi-Fi RSSI datasets, TIE1 and SAH1, were collected at Tampere University buildings in late 2017 for indoor positioning research. Each dataset provides separate training and test files containing Received Signal Strength Indicator (RSSI) matrices from hundreds of access points paired with ground-truth 3D coordinates and floor labels. The data includes a specific placeholder value (+100dBm) for non-detected access points, a common preprocessing step for fingerprinting algorithms.

Use Cases

• Training fingerprint-based localization models using the radio map matrices (training_rss) and coordinate labels (training_coordinates).
• Evaluating model performance on separate operational fingerprint data (test_rss) with known ground truth locations (test_coordinates).
• Studying signal propagation and access point density variations between two distinct campus buildings (Tietotalo and Sahkotalo).
• Preprocessing experiments involving the imputation of the default +100dBm value for non-detected access points.

Strengths

• Provides a clear train-test split with 10,633 training and 50 test samples for TIE1, and 9,291 training and 156 test samples for SAH1.
• Includes precise 3D ground truth coordinates (X, Y, Z, floor) for all reference and evaluation samples.
• Covers a substantial number of Wi-Fi access points: 613 in one building and 775 in another, suggesting detailed radio environment mapping.
• Documentation explicitly describes the data collection method (regular smartphone) and the handling of missing AP signals.

Limitations

• The data is from 2017, which may limit its relevance for modeling contemporary Wi-Fi environments or newer standards.
• Coordinates are in a local building system, not a global system like WGS84, requiring transformation for some applications.
• Metadata such as exact update dates, file sizes, and detailed column descriptions are not provided across sources.

Provenance

Source

Tampere University (author: Elena Simona Lohan)

Collection Method

Collected using a regular smartphone, storing raw RSSI values; non-detected access points were filled with a placeholder value (+100dBm).

Time Range

August-December 2017

Geography

Tietotalo and Sahkotalo buildings, Tampere University campus, Finland