Multimodal Dataset for Wireless Channel ML

Dataset v1.0 Open Access (Registration Required) Machine Learning Ready

Multimodal Dataset for Wireless Channel Machine Learning

A richly annotated multimodal dataset combining RF channel measurements, environmental images, and location data — designed to enable machine learning research for channel estimation, LoS/NLoS classification, and environment-aware radio resource management.

Download Dataset Learn More
3
Modalities
—
Measurement Points
—
Scenarios
Sub-6 GHz
Frequency
LoS / NLoS
Propagation Labels
Python / MATLAB
Sample Code

Data Modalities

RF Channel Data

Raw channel transfer functions and derived channel parameters including path loss, delay spread, angular spread, and Rician K-factor measured with a MIMO channel sounder.

.mat .csv
Environmental Images

360° panoramic images captured at each receiver location, providing visual context of the propagation environment for vision-aided channel prediction and scene classification.

.jpg .png
Location & Map Data

GPS coordinates (latitude, longitude, altitude) for each measurement point along with building geometry data for spatial analysis and environment-aware modeling.

.csv .geojson

Supported ML Tasks

LoS / NLoS Classification Path Loss Prediction Channel Parameter Estimation Beam Selection & Management Environment-Aware Radio Resource Management Multimodal Feature Fusion

Measurement Scenarios

Scenario Environment Frequency Measurement Points LoS / NLoS Label Image Available
Scenario A Urban Outdoor (Macro) 4.85 GHz — Yes Yes
Scenario B Urban Outdoor (Micro) 4.85 GHz — Yes Yes
Scenario C Indoor (Office) — — TBD TBD

Available Files

File Name Contents Format Size
multimodal_full_v1.zip All modalities (RF + Images + GPS) Mixed — GB Requires registration
rf_channel_data.zip RF channel measurements only .mat / .csv — MB Requires registration
images_360.zip 360° panoramic images at all RX points .jpg — GB Requires registration
location_data.zip GPS coordinates and building geometry .csv / .geojson — MB Requires registration
sample_code.zip Python & MATLAB data loading scripts .py / .m — KB Requires registration
README.txt Dataset structure, field descriptions, usage guide .txt — KB Requires registration

Download

This dataset is freely available for academic and research purposes. Please complete the registration form below. The download link will be provided immediately after submission.

Citation

If you use this dataset in your research, please cite the following: 

@dataset{kim2024multimodal,
  author      = {Kim, Minseok and others},
  title       = {{Multimodal Dataset for Wireless
                 Channel Machine Learning}},
  institution = {Niigata University},
  year        = {2024},
  url         = {https://radio.eng.niigata-u.ac.jp/}
}

Related Publications

  • IEEE Access 2021
    M. Kim, S. Tang, K. Kumakura, “Fast Double-Directional Full Azimuth Sweep Channel Sounder,” IEEE Access, vol. 9, pp. 80288–80299, Jun. 2021.
    DOI Link
  • IEEE Access 2023
    A. Ghosh, M. Kim, “THz Channel Sounding and Modeling Techniques: An Overview,” IEEE Access, vol. 11, pp. 17823–17856, Feb. 2023.
    DOI Link

License & Terms of Use

This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

  • Free to use for academic and non-commercial research
  • Redistribution requires attribution to the original source
  • Commercial use requires prior written permission
  • The dataset is provided “as is” without warranty of any kind
  • Do not attempt to re-identify individuals from image data

Contact

For questions regarding the dataset or collaboration inquiries, please contact:

Prof. Minseok Kim
Radio Signal Processing Laboratory
Graduate School of Science and Technology
Niigata University, Japan
mskim@eng.niigata-u.ac.jp
radio.eng.niigata-u.ac.jp