MVRL/dense500-retrieval

Dense500 — retrieval split

Satellite tile imagery for hierarchical object-to-tile retrieval. Each site is a complete tile pyramid: every 256×256 tile at every zoom level between the site's root and its leaves. This split is the evaluation gallery of GeoDIVE (WACV 2026) and belongs to the Dense500 collection introduced with Genesis (ACM SIGSPATIAL 2026). Object annotations and query captions are in MVRL/dense500-retrieval-instances.

Construction

Site centers were sampled worldwide with a population prior and continent stratification. For each site, every tile between its root zoom and its leaf zoom was collected, so each site forms a full quadtree. Sites whose imagery source changes between adjacent zoom levels are flagged in the schema rather than removed.

Schema

Each row is one tile.

Column	Type	Description
image	image	256×256 JPEG
location	string	Tile key "{zoom}_{x}_{y}" (Web Mercator)
site_id	string	Groups all tiles of one quadtree
tag	string	naeu_urban, global_urban, or rural
continent	string	NA, EU, Asia, Africa, SA, Oceania
site_lon, site_lat	float64	Site center (degrees)
zoom, tile_x, tile_y	int32	Web Mercator tile coordinates
site_start_zoom, site_leaf_zoom	int32	Root and leaf zoom of the site
site_depth	int32	leaf - start + 1
is_root, is_leaf	bool	Tile position in its site
site_has_source_shift	bool	Site shows a source change between adjacent zooms
site_shift_zoom	int32	First zoom after the change, -1 if none
is_post_shift	bool	zoom >= site_shift_zoom when a change exists

Loading

from datasets import load_dataset

ds = load_dataset("MVRL/dense500-retrieval", split="train")

from collections import defaultdict
by_site = defaultdict(list)
for row in ds:
    by_site[row["site_id"]].append(row)

Annotations join on the location column.

Citation

@inproceedings{iqbal2026geodive,
  author    = {Iqbal, Hamza and Cher, Dan and Xing, Eric and Kolouju, Pranavi and Stylianou, Abby and Pless, Robert and Jacobs, Nathan},
  title     = {GeoDIVE: Geospatial Descent Informed by Value Estimation},
  booktitle = {IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
  year      = {2026}
}

@inproceedings{khanal2026genesis,
  author    = {Khanal, Subash and Cui, Yangzhi and Cher, Daniel and Xing, Eric and Wei, Brian and Sastry, Srikumar and Jacobs, Nathan},
  title     = {Genesis: A Generative Engine for Hierarchical Satellite Image Synthesis},
  booktitle = {ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (ACM SIGSPATIAL)},
  year      = {2026},
  doi       = {10.1145/3841645.3843313}
}