Simantiris et al. (2025) AIFloodSense

Identification

• Journal: Mendeley Data
• Year: 2025
• Date: 2025-12-25
• Authors: Georgios Simantiris, Konstantinos Bacharidis, Apostolos Papanikolaou, Petros Giannakakis, Costas Panagiotakis
• DOI: 10.17632/4p4rcg8jz4

Research Groups

Georgios Simantiris, Konstantinos Bacharidis, Apostolos Papanikolaou, Petros Giannakakis, Costas Panagiotakis

Short Summary

AIFloodSense is a novel, multi-modal aerial imagery dataset designed to advance machine learning for flood classification, semantic segmentation, and scene understanding, featuring high-resolution images from 230 global flood events between 2022 and 2024.

Objective

• To create a comprehensive, multi-modal aerial imagery dataset (AIFloodSense) to support and benchmark machine learning models for classification, semantic segmentation, and scene understanding in diverse flooded environments.

Study Configuration

• Spatial Scale: Global coverage across 6 continents and 64 countries, encompassing urban, peri-urban, and rural environments with high-resolution aerial imagery.
• Temporal Scale: Data collected from 230 distinct flood events occurring between 2022 and 2024.

Methodology and Data

• Models used: Not applicable; this paper describes a dataset designed for use with models like deep learning for classification, semantic segmentation, and scene understanding.
• Data sources: High-resolution aerial images primarily captured by Unmanned Aerial Vehicles (UAVs).

Main Results

• The AIFloodSense dataset was developed, comprising multi-modal, high-resolution aerial imagery.
• The dataset covers 6 continents, 64 countries, and 230 flood events from 2022 to 2024.
• It includes diverse environmental contexts (urban, peri-urban, rural) and camera angles (sky absence, sky presence).
• AIFloodSense is specifically designed to serve as a benchmark and training resource for classification, semantic segmentation, and scene understanding tasks in flooded environments.

Contributions

• Introduction of AIFloodSense, a novel, multi-modal, high-resolution aerial imagery dataset, addressing a critical need for comprehensive flood-related machine learning data.
• Provides global coverage and diverse environmental contexts, enhancing the generalizability of models trained on it.
• Facilitates benchmarking and development of advanced deep learning models for flood classification, semantic segmentation, and scene understanding.

Funding

• Not specified in the provided text.

Citation

@article{Simantiris2025AIFloodSense,
  author = {Simantiris, Georgios and Bacharidis, Konstantinos and Papanikolaou, Apostolos and Giannakakis, Petros and Panagiotakis, Costas},
  title = {AIFloodSense},
  journal = {Mendeley Data},
  year = {2025},
  doi = {10.17632/4p4rcg8jz4},
  url = {https://doi.org/10.17632/4p4rcg8jz4}
}