M4Heights: Benchmark Dataset for Building Height Estimation

📻 Overview

This is the official dataset repository for M4Heights dataset, a multi-modal, multi-resolution, and multi-temporal dataset designed for building height estimation. The dataset contains approximately 1 million images comprising time series of Sentinel-1 SAR and Sentinel-2 MSI satellite data, high-resolution aerial orthophotos, and high-quality building height reference maps. Additionally, M4Heights provides the largest associated multi-image super-resolution dataset to enhance height estimation accuracy. Our open-source dataset supports a range of modeling approaches, offering extensibility to new geographic regions and advancing the development of DL models for building height estimation.

🚀 Usage

We have tested the dataset on several deep learning models. The code for the models, dataset usage along with the instructions are provided on our Github Repository. For more details on dataset and experiments, please check our paper.

Codebase : https://github.com/RituYadav92/M4Heights

Paper : https://www.nature.com/articles/s41597-025-06495-3

⬇️ Data Size:

  dataset (zip)    |    size (325 GB)
  ---------------------------------
  Train            |      ≈275.0GB
  Test             |       ≈48.0GB
  Sample Data      |       120.0MB

  * Unzipped dataset size is ≈900GB

🎯 Stats and Samples

🎓 Citation:

@misc {ritu_yadav_2025,
    author       = { {Ritu Yadav} },
    title        = { M4Heights },
    year         = 2025,
    url          = { https://huggingface.co/datasets/Rituxx96x/M4Heights },
    doi          = { 10.57967/hf/4765 },
    publisher    = { Hugging Face }
}

🎓 Dataset Paper Citation:

Yadav, R., Nascetti, A. & Ban, Y. A Multi-Modal, Multi-Temporal, Multi-Resolution Benchmark Dataset for Building Height Estimation. Sci Data (2025). https://doi.org/10.1038/s41597-025-06495-3

👋 Contact Info.:

Ritu Yadav (email: [email protected])