MagWi: Benchmark dataset for long term magnetic field and Wi-Fi data involving heterogeneous smartphones, multiple orientations, spatial diversity and multi-floor buildings

Citation Author(s):
Information and Communication Engineering, Yeungnam University
Information and Communication Engineering, Yeungnam University
Muhammad Usman
Department of Computer Science, University of Gujrat
Information and Communication Engineering, Yeungnam University
Yousaf Bin
Information and Communication Engineering, Yeungnam University
Information and Communication Engineering, Yeungnam University
Submitted by:
Imran Ashraf
Last updated:
Tue, 05/17/2022 - 22:21
Data Format:
Link to Paper:
2 ratings - Please login to submit your rating.


This study presents a dataset that comprises the magnetic field, Wi-Fi, and the data from the inertial measurement unit (IMU) sensors of the smartphone including accelerometer, gyroscope, and barometer. First, the important

characteristics of both the Wi-Fi and the magnetic field that require further investigation are highlighted, and later the data are collected. The data are collected over a longer period spanning approximately five years involving five

different smartphones used by four different users, both female, and males. Different path geometries are followed in different multi-floor buildings which are physically separated, comprising both small and large areas. Besides,

three different orientations of the smartphone are considered for data collection covering corridors, halls, and laboratories. The data from the stairs help to test ’stairs up’ and ’stairs down’ events and approaches aiming at multi-floor

positioning can be tested with the provided dataset.


For instructions, kindly check the following paper which is published in IEEE Access:

"MagWi: Benchmark Dataset for Long Term Magnetic Field And Wi-Fi Data Involving Heterogeneous Smartphones, Multiple Orientations, Spatial Diversity and Multi-floor Buildings".

For further queries, contact at


The Wi-Fi part of this dataset appears to have multiple issues:
1. Filenames end with .csv, however they are actually .xls files. (Files in IMU / geomagnetic part are correctly labelled.)
2. Every location record under "BE Engineering", "CS Engineering", "Electrical Engineering", "IACT" (called RIC building in the paper) and "IT Engineering" has a Y position of 24, with changes only in X. This contradicts the track given in the paper. The X/Y positions in files with same timestamp under "Magnetic field dataset" appear to be more plausible.
3. Not every scenario present in the dataset is covered in the paper, for example the track for Scenario 3 of IT Engineering is not mentioned. Also, data for Scenario 1 of EE and RIC / IACT Building are not present in the dataset.

It should also be noted that 'Magnetic field dataset\Continuous Data\IT Engineering\Navigation\Scenario 2' contains no data file.

Submitted by Tianshu Zhang on Mon, 04/04/2022 - 06:32

I found the same problem as the one on the first floor, and look forward to the reply on the dataset.

Submitted by James Owen on Thu, 11/03/2022 - 10:15

No clear evidence shows how the authors divided the training and testing set to get the positioning results.

Submitted by YINHUAN DONG on Fri, 03/03/2023 - 04:41