datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ai4g-flood-dataset
Flood Detection Dataset
Introduction
This dataset accompanies the paper Mapping global floods with 10 years of satellite radar data (Nature Communications, 2025) and contains global flood detections derived from Sentinel-1 Synthetic Aperture Radar (SAR) imagery using a deep learning change detection model. The dataset spans October 2014 – September 2024, offering a longitudinal view of flood-prone areas worldwide.
Key features:
Cloud-penetrating SAR data for consistent… See the full description on the dataset page: https://huggingface.co/datasets/ai-for-good-lab/ai4g-flood-dataset.floodnetRehosted from Google drive from Paper Repo.
If you use this dataset, please cite:
https://arxiv.org/abs/2012.02951
FloodNet_2021-Track_2_Dataset_HF
FloodNet: High Resolution Aerial Imagery Dataset for Post-Flood Scene Understanding
This is the HF-hosted version of FloodNet.
The FloodNet 2021: A High Resolution Aerial Imagery Dataset for Post-Flood Scene Understanding provides high-resolution UAS imageries with detailed semantic annotation regarding the damages. To advance the damage assessment process for post-disaster scenarios, the authors of the dataset presented a unique challenge considering classification, semantic… See the full description on the dataset page: https://huggingface.co/datasets/takara-ai/FloodNet_2021-Track_2_Dataset_HF.floodcastbench
FloodCastBench (mirror)
Re-hosted for use in a CNN-LSTM spatiotemporal flood forecasting course project, to allow
selective per-region download from Google Colab (snapshot_download(..., allow_patterns=...))
instead of a single non-resumable 21.6GB Zenodo archive.
Original dataset: Xu, Q., Shi, Y., Zhao, J. et al. "FloodCastBench: A Large-Scale Dataset and
Foundation Models for Flood Modeling and Forecasting." Scientific Data 12, 431 (2025).… See the full description on the dataset page: https://huggingface.co/datasets/mahajananhad/floodcastbench.FSCM_Flood_nerffloodpulse-sft-v1
FloodPulse SFT
Temporal Sentinel-2 pairs for flood-relevant events. MNDWI water masks on pre/post scenes define new inundation (water gained); tiles get a caption and optional grounding rows for inundation regions.
Record counts (this build)
Split
JSONL lines
train
263
validation
31
test
0
total
294
Tiles processed (caption rows ≈ this; grounding rows added when regions exist): 189.
Inputs
Events: JSON/CSV with event_id, lat, lon… See the full description on the dataset page: https://huggingface.co/datasets/NuTonic/floodpulse-sft-v1.FSCM_Flood_playgroundFloodNet-Challenge-EARTHVISION2021-Track1flash-flood-benchmark-figures
TORRENT — portal asset backend
Rendered images and download packages for the TORRENT flash-flood benchmark.
This repository is not a standalone dataset: it is the static asset backend that
the TORRENT portal Space loads images and zip bundles from over the Hub CDN.
Portal (Space): https://huggingface.co/spaces/skyan1002/flash-flood-benchmark
Dataset (data of record on the Hub): https://huggingface.co/datasets/skyan1002/flash-flood-benchmark-data
Archive of record (Zenodo v1.0):… See the full description on the dataset page: https://huggingface.co/datasets/skyan1002/flash-flood-benchmark-figures.flood-detection-pair-colombia
flood-detection-pair-colombia
110 paired Sentinel-2 satellite tile samples (4 PNGs per row) labeled for flood detection across 9 documented flood events in Colombia. Built for fine-tuning a small vision-language model that runs on a satellite or community ground station — see the humaid project.
What's in each sample
Each row in flood_train.jsonl / flood_eval.jsonl is a 4-image vlm_sft example. The user message contains four image content blocks (in this order) followed… See the full description on the dataset page: https://huggingface.co/datasets/jpmarindiaz/flood-detection-pair-colombia.dataFSCM_Flood_Kittiafrica_flood_occurrence
Africa Flood Occurrence
Annual flood occurrence across Africa at ~500 m, 2001 to 2024, from two parallel IWMI series on one grid: flood_occurrence (the finer count, per-year maximum 17 to 37) and inter_annual_flood_occurrence (consistently smaller). Both start at 1 and neither is documented by the publisher, but their distributions are those of counts. Rebuilt from IWMI Africa GeoPortal tile services, which serve no downloadable raster.
Region Africa
Period 2001-2024
Theme… See the full description on the dataset page: https://huggingface.co/datasets/IWMIHQ/africa_flood_occurrence.africa-synth-displacement-flooding-displacement-health-all
Flooding & Displacement Health | Africa (Electric Sheep Africa metadata inventory)
Size category: 10K<n<100K - Formats: csv - Sector: climate_environment - Engineered by Electric Sheep Africa
TL;DR
This dataset is part of the Electric Sheep Africa catalog on Hugging Face. It is indexed for African data discovery with standardized metadata, loading guidance, provenance notes, and analyst-oriented context.
What This Dataset Covers
Public datasets… See the full description on the dataset page: https://huggingface.co/datasets/electricsheepafrica/africa-synth-displacement-flooding-displacement-health-all.d2p_dataset
The D2P dataset
The D2P dataset is a dataset based on the Depth2Pose monocular depth estimation benchmark, a pose-based evaluation of MDEs without ground-truth depth. The dataset contains challenging scenes beyond the distribution of common training data, together with a simple and extensible evaluation framework, presented on the github page. The scenes are divided into two categories: statues and vegetation. Undistorted images and reconstructions in standard colmap format is… See the full description on the dataset page: https://huggingface.co/datasets/floodgab/d2p_dataset.africa-synth-flooding-extreme-weather-trauma-all
Extreme Weather & Trauma (SSA) | Africa (Electric Sheep Africa metadata inventory)
Size category: 10K<n<100K - Formats: csv - Sector: climate_environment - Engineered by Electric Sheep Africa
TL;DR
This dataset is part of the Electric Sheep Africa catalog on Hugging Face. It is indexed for African data discovery with standardized metadata, loading guidance, provenance notes, and analyst-oriented context.
What This Dataset Covers
Public datasets… See the full description on the dataset page: https://huggingface.co/datasets/electricsheepafrica/africa-synth-flooding-extreme-weather-trauma-all.africa-synth-flooding-property-insurance-risk-all
African Property Insurance Risk | Africa (Electric Sheep Africa metadata inventory)
Size category: 100K<n<1M - Formats: csv - Sector: climate_environment - Engineered by Electric Sheep Africa
TL;DR
This dataset is part of the Electric Sheep Africa catalog on Hugging Face. It is indexed for African data discovery with standardized metadata, loading guidance, provenance notes, and analyst-oriented context.
What This Dataset Covers
Public datasets… See the full description on the dataset page: https://huggingface.co/datasets/electricsheepafrica/africa-synth-flooding-property-insurance-risk-all.ETCI-2021-Flood-Detection
ETCI 2021 Flood Detection Dataset
Description
The ETCI 2021 Flood Detection Dataset is a comprehensive flood detection segmentation dataset that focuses on SAR (Synthetic Aperture Radar) images taken by the ESA Sentinel-1 satellite. This dataset provides pairs of VV (Vertical Transmit, Vertical Receive) and VH (Vertical Transmit, Horizontal Receive) polarization images, which have been processed by the Hybrid Pluggable Processing Pipeline (hyp3). Additionally… See the full description on the dataset page: https://huggingface.co/datasets/luisrH/ETCI-2021-Flood-Detection.SDSU_MidWest_Flood_2019
SDSU Midwest Flood Dataset 2019
This dataset provides a benchmark for flood detection in satellite imagery. It contains true-color satellite images from the 2019 Midwest USA flooding event, with each image paired with a binary mask indicating flooded areas.
Important Update
This dataset has been revised to provide segmentation-ready image-mask pairs through the Hugging Face datasets library.
Each sample now contains:
image: RGB satellite image
mask: binary flood… See the full description on the dataset page: https://huggingface.co/datasets/youngsun05/SDSU_MidWest_Flood_2019.d2p_dataset_example
The D2P dataset
The D2P dataset is a dataset based on the Depth2Pose monocular depth estimation benchmark, a pose-based evaluation of MDEs without ground-truth depth. The dataset contains challenging scenes beyond the distribution of common training data, together with a simple and extensible evaluation framework, presented on the github page. The scenes are divided into two categories: statues and vegetation. Undistorted images and reconstructions in standard colmap format is… See the full description on the dataset page: https://huggingface.co/datasets/floodgab/d2p_dataset_example.africa-synth-flooding-leptospirosis-environmental-all
Leptospirosis Environmental Surveillance | Africa (Electric Sheep Africa metadata inventory)
Size category: 10K<n<100K - Formats: csv - Sector: health - Engineered by Electric Sheep Africa
TL;DR
This dataset is part of the Electric Sheep Africa catalog on Hugging Face. It is indexed for African data discovery with standardized metadata, loading guidance, provenance notes, and analyst-oriented context.
What This Dataset Covers
Health datasets… See the full description on the dataset page: https://huggingface.co/datasets/electricsheepafrica/africa-synth-flooding-leptospirosis-environmental-all.flood_segmentationafrica-synth-flooding-sea-level-rise-coastal-health-all
Sea-Level Rise & Coastal Health Infrastructure (SSA) | Africa (Electric Sheep Africa metadata inventory)
Size category: 10K<n<100K - Formats: csv - Sector: health - Engineered by Electric Sheep Africa
TL;DR
This dataset is part of the Electric Sheep Africa catalog on Hugging Face. It is indexed for African data discovery with standardized metadata, loading guidance, provenance notes, and analyst-oriented context.
What This Dataset Covers
Health… See the full description on the dataset page: https://huggingface.co/datasets/electricsheepafrica/africa-synth-flooding-sea-level-rise-coastal-health-all.flood-binary-hnm-benchmark
Flood Binary HNM Benchmark
Street-level flood/non-flood binary classification imagery, used in "Improving CRIS-HAZARD: Automated First-Pass Flood Image Screening via Phase-1 Hard Negative Mining" (Singh & Dixon, pending submission to Computers & Geosciences).
4,099 deduplicated (SHA-256 exact-match only, no perceptual dedup) street-level images, stratified 80/20 by fine-grained category, seed=42. Flood prevalence is 39.3% in both splits. No held-out test split — the paper… See the full description on the dataset page: https://huggingface.co/datasets/zinnia82/flood-binary-hnm-benchmark.FSCM_Flood_trainFloodFloodNet_VQAfloodsflood-susceptibility-mapFSCM_Flood_kitti_more_less