786 color clusters contain satellite image tiles harvested from ArcGIS World Imagery. Each tile is a 256x256 PNG snapshot at zoom level 15, organized using K-Means centroids in CIE LAB space. The dataset was created by Daksh17440 and last updated on 2026-05-17.
Use Cases
- Training image segmentation models based on perceptually-uniform color clusters.
- Developing content-based image retrieval systems for satellite imagery.
- Analyzing land cover patterns based on color-space groupings.
- Benchmarking clustering algorithms on high-dimensional image data.
- Creating synthetic training data by sampling from distinct color buckets.
Strengths
- Tiles are organized into 5,000 target color clusters using K-Means in LAB space.
- Each tile is a 256x256 PNG at a consistent zoom level 15 (~4.8 m/pixel).
- Data is deduplicated and sourced from ArcGIS World Imagery.
Limitations
- Dataset coverage is partial, with tiles for only 786 out of the target 5,000 clusters.
- Column-level documentation and sample data are unavailable.
- The geographic and temporal coverage of the source imagery is unspecified.
Provenance
- Source
- ArcGIS World Imagery
- Collection Method
- Harvested and processed using the Mosaic Library, then clustered with K-Means in CIE LAB space.
- Freshness
- Last updated 2026-05-17 19:56:24.