Sign in to view source links and access this dataset
Description
22,926,427 instance masks of individual agricultural fields are provided across 672,909 high-resolution satellite image patches. The dataset, created by MykolaL, excludes Pleiades satellite imagery due to third-party policies but maintains broad coverage. It was last updated on August 22, 2025.
Use Cases
Training instance segmentation models based on the 22 million field masks.
Developing multi-resolution analysis pipelines based on the 0.25m to 10m image patches.
Conducting agricultural land-use studies based on the large-scale field boundary data.
Benchmarking geospatial computer vision models on a dataset of over 600k image patches.
Strengths
Contains 22,926,427 instance masks, providing a large-scale training resource.
Comprises 672,909 high-resolution image patches for diverse model input.
Covers a multi-resolution range from 0.25 meters to 10 meters per pixel.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
The dataset structure is split into multiple archives, which may complicate initial access.
Provenance
Source
huggingface
Collection Method
Likely compiled from multiple satellite missions, with Pleiades data excluded.
Time Range
null
Freshness
Last updated 2025-08-22 09:23:17; freshness should be verified.
Geography
null
Pleiades satellite mission imagery is excluded due to third-party usage policies.