datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
us_wildfires_stpp
US Wildfires STPP Benchmark Dataset
A benchmark-ready Spatio-Temporal Point Process (STPP) dataset derived from 1.88 Million US Wildfires, following the standard split semantics for Neural STPP evaluation.
Dataset Description
Each record represents a sequence of events. The dataset covers historical wildfire incidents across the US, partitioned sequentially into train / val / test subsets (70% / 15% / 15% ratio).
Source Format
Raw data was obtained… See the full description on the dataset page: https://huggingface.co/datasets/seahorse-stpp/us_wildfires_stpp.wildfire-korea-episodes-300m
South Korean Wildfire Episodes Dataset (300 m Grid)
A geospatial dataset of South Korean wildfire progression combining NASA FIRMS (VIIRS/MODIS) fire detections with static environmental layers (DEM, landcover, slope, NDVI, forest structure, etc.).
Tiles are preprocessed into a 300 m spatial grid and segmented into temporal wildfire episodes suitable for reinforcement-learning–based wildfire spread prediction research.
Overview
This dataset provides a unified, model… See the full description on the dataset page: https://huggingface.co/datasets/chaseungjoon/wildfire-korea-episodes-300m.