Dataset for Radio Frequency Based Device-Free Crowd Size Estimation in Public Transportation Environments

This dataset was collected with the aim to evaluate the use of device-free wireless sensing to perform crowd size estimation in public transportation environments. This dataset contains measured Received Signal Strength Indicator (RSSI) data and manual people count data from 2 environments. The first environment is a platform of a subway station in Antwerp, Belgium. We will refer to this environment as the indoor environment. The second environment is an outdoor platform in Barcelona, Spain. We will refer to this environment as the outdoor environment.

The measurement RSSI data is gathered using a wireless sensor network performing a measurement cycle every 10 seconds in which the sensor nodes sequentially transmit a message which is received by the other nodes in the network, storing the received signal strength upon reception. The message payload contains the list of last collected RSSI values for each node. This payload is received by the gateway and forwarded to the back-end server for storage and processing. Each of these messages as received by the gateway are represented by a single line in the raw RSSI data as found in this dataset.

The ground truth data has 2 types of values: values greater or equal to zero are people counts, negative values represent a rail vehicle entering (-1) or leaving (-2) the environment. These rail vehicle event values can be used to segment the dataset in counts when the rail vehicle was present or not, or for building a classification model to detect the presence of rail vehicles.

The node locations are stored in JSON format. For the outdoor environment these are geographic coordinates in GeoJSON format. For the indoor environment these are relative indoor coordinates in meter.

Indoor Dataset

This dataset is comprised of 4 full days of RSSI data and 3 instances of collected ground truth data. This data was collected at the Groenplaats subway station in the city of Antwerp, Belgium.

Dataset file structure:

data_platform_indoor
├── wsn_indoor_environment.json
├── rssi_data
│   ├── rssi_platform_indoor_2023-05-17.csv
│   ├── rssi_platform_indoor_2023-05-18.csv
│   ├── rssi_platform_indoor_2023-05-19.csv
│   └── rssi_platform_indoor_2023-07-17.csv
└── training_data
    ├── training_platform_indoor_2023-05-17.csv
    ├── training_platform_indoor_2023-05-19.csv
    └── training_platform_indoor_2023-07-17.csv

Notes on the data:

2023-05-17 : training data available, batteries replaced that day -> no calibration possible
2023-05-18 : can be used for calibration of 2023-05-17
2023-05-19 : training data available
2023-07-17 : training data available

Outdoor Dataset

This dataset is comprised of 3 full days of Received Signal Strength Indicator (RSSI) data and 3 instances of collected ground truth data. This data was collected at the Hospital Sant Joan Despí | TV3 tram stop in Barcelona, Spain.

Dataset file structure:

data_platform_outdoor
├── wsn_outdoor_environment.geojson
├── rssi_data
│   ├── rssi_platform_outdoor_2024-06-11.csv
│   ├── rssi_platform_outdoor_2024-06-12.csv
│   └── rssi_platform_outdoor_2024-07-02.csv
└── training_data
    ├── training_platform_outdoor_2024-06-11.csv
    ├── training_platform_outdoor_2024-06-12.csv
    └── training_platform_outdoor_2024-07-02.csv

Notes on the data:

2024-06-11 : training data available, batteries replaced that day -> no calibration possible
2024-06-12 : training data available
2024-07-02 : training data available, training data gathered by other person than previous training data, no people counts during stopped trams