datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
USA-Map-Tiles
Dataset Description
The dataset is a set of map tiles for the various states of the USA.
License: Open Data Commons Open Database License (ODbL) v1.0
Dataset Sources
OpenStreetMap.org
Geofabrik Download Server
Bounding Boxes per State
Dataset Structure
Organized into folders by state
Structure of graphml files:
Top of files have a list of keys/ids that correspond to the properties of the segment:
maxspeed: speed limit
oneway: if it is a… See the full description on the dataset page: https://huggingface.co/datasets/incognitolm/USA-Map-Tiles.mahjong_souls_tiles
Dataset Card for Mahjong Souls Tiles
Dataset Details
Dataset Description
This dataset, pjura/mahjong_souls_tiles, contains images of Mahjong tiles extracted from the game Mahjong Soul. It was created specifically for the purpose of training computer vision models to recognize Mahjong tiles within the game interface.
The dataset was used to fine-tune the google/vit-base-patch16-224-in21k model, resulting in the pjura/mahjong_vision model available on Hugging… See the full description on the dataset page: https://huggingface.co/datasets/pjura/mahjong_souls_tiles.haidian-building-tilespixelrag-tiles
PixelRAG tile corpus
Rendered screenshot tiles for PixelRAG, a visual retrieval-augmented-generation system that retrieves over page images instead of parsed text. Each Wikipedia page is rendered to an image and cut into fixed-height tiles; retrieval runs on the tiles directly with a Qwen3-VL embedding model.
This repository holds the full tile corpus that the published FAISS indexes and embeddings were built from, so the whole pipeline (tiles → embeddings → index → search) can… See the full description on the dataset page: https://huggingface.co/datasets/StarTrail-org/pixelrag-tiles.win-tiles-icons
Dataset Details
Dataset Description
The pictures were taken from the Discord server https://discord.gg/VMz3GD4d
Relevance as of 29.11.2024
I tried to classify some parts, but I have clumsy and crooked paws to make a proper classifier for all this.
plism-dataset-tiles
PLISM dataset
The Pathology Images of Scanners and Mobilephones (PLISM) dataset was created by (Ochi et al., 2024) for the evaluation of AI models’ robustness to inter-institutional domain shifts.
All histopathological specimens used in creating the PLISM dataset were sourced from patients who were diagnosed and underwent surgery at the University of Tokyo Hospital between 1955 and 2018.
PLISM-wsi consists in a group of consecutive slides digitized under 7 different scanners and… See the full description on the dataset page: https://huggingface.co/datasets/owkin/plism-dataset-tiles.curbai-tilesnanopath-fairness-tiles
nanopath-fairness-tiles
Pre-tiled histopathology patches from CPTAC whole-slide images, used as the
external out-of-distribution validation set for a study on pretraining-time
vs. post-hoc fairness in histopathology foundation models.
Contents
Per-cohort folders, each slides_full/<slide_id>.parquet (one row per tile:
case_id, slide_id, tile_idx, image) + labels.tsv:
cohort
organ / task
slides
cptac_lung
NSCLC — LUAD vs LSCC subtype
604
cptac_gbm
GBM —… See the full description on the dataset page: https://huggingface.co/datasets/ryankim17920/nanopath-fairness-tiles.global-openstreetmap-extraction-slippy-tiles-tarurban-opengen-tiles-v2
Urban OpenGen tiles v2: sixty cities from OpenStreetMap
Sixty cities on six continents, each a 10 km × 10 km window centred on the core, cut into
a 25 × 25 grid. 625 tiles of 400 m per city, 37,500 in all, of which 30,957 carry built
content; the rest are sea, or ground OpenStreetMap has nothing for. Every tile is
128 × 128 px at 3.125 m per pixel.
Every city was rendered by the same code with the same conventions, which is the whole
point. Opportunistic crops from different… See the full description on the dataset page: https://huggingface.co/datasets/differential-studio/urban-opengen-tiles-v2.onlypits-lunar-tilesglobal-elevation-aster-slippy-tiles-tar-gz
PeakNav global elevation tiles
Slippy-map elevation tiles (zoom 8, ten resolutions each, packed per zoom-6
tile as elev_tiles/zoom_06/x_*/y_*.tar.gz) used by the PeakNav app.
Sources and attribution
ASTER GDEM v3 (worldwide base): ASTER GDEM is a product of METI and
NASA.
OpenStreetMap summit elevations (natural=peak / ele): © OpenStreetMap
contributors, ODbL.
swissALTI3D (Switzerland and Liechtenstein): ©swisstopo - Federal Office
of Topography swisstopo, free… See the full description on the dataset page: https://huggingface.co/datasets/PeakNav/global-elevation-aster-slippy-tiles-tar-gz.tilebac-ULDM-tiles
Background
Overview
This dataset contains ultralow-dose cryoEM montage tile images of the bacteria Pantoea sp. YR343. Segmentation models from YOLOv11, YOLO26, U-Net, Detectron2 and SAM3 have been fine-tuned to predict bacterial inner membranes and outer membranes. This bacterial membrane dataset is a benchmark dataset to challenge current AI workflows in rapid, seamless montage stitching and stitched segmentation in extremely noisy ultralow-dose cryoEM images. Bacterial flagella low-dose… See the full description on the dataset page: https://huggingface.co/datasets/LynnMass/tilebac-ULDM-tiles.gtex-10M-balanced-tiles
GTEx 10M Balanced Tiles
This dataset contains 10,000,000 JPEG-encoded 224x224 pathology tiles from GTEx SVS slides in s3://path-datasets/gtex/svs_by_tissue, balanced at 250,000 tiles for each of the 40 tissue prefixes. Source objects under 80,000,000 bytes are ignored because the GTEx prefix contains tiny unsupported SVS objects. Tiles are sampled from virtual OpenSlide levels 0, 1, and 2, so the effective field of view varies while the emitted JPEG size stays fixed.
Each… See the full description on the dataset page: https://huggingface.co/datasets/medarc/gtex-10M-balanced-tiles.game-asset-tiles
Game Asset Tiles
This dataset contains 275 samples, each consisting of a tile image, a matching template image and a description.
Tiles were generated with Nano Banana 2 and FLUX.2, or sourced from kenney.nl.
hex / oct / rect describe the footprint shape, flat-top or pointy-top describe the hex shape orientation, and isometric, oblique, or top-down describe the perspective.
Repository layout:
tiles/: PNG tile images
descriptions/: Short English text prompts
coordinates/: JSON point… See the full description on the dataset page: https://huggingface.co/datasets/crafiq/game-asset-tiles.office-flight-tracker-london-tiles
London Streets.gl tiles
Generated 2026-09-17T12:29:11.517Z from © OpenStreetMap contributors.
Format: uncompressed Mapbox Vector Tiles generated by the
StrandedKitty/planetiler StreetsProfile, zooms 9–16.
mahjong_tilesdaesang-tiles
daesang-tiles
스페이스팜 대상지도 대시보드용 지도 타일. buildings_kr.pmtiles: 국토교통부 V-World GIS건물통합정보(기준일 2026-09-09, CC BY) 전국 16개 시도 → PMTiles(z12~15). 속성 pnu·name·use·floors·height(m; 높이>0 → 높이, 아니면 층수×3.3, 없으면 4).
pf-grid-tiles
High Resolution Precipitation Climatologies from NASA Precipitation Measurement Missions
Author: Stephen W. Nesbitt, Department of Climate, Meteorology & Atmospheric
Sciences, University of Illinois Urbana-Champaign.
Annual, 0.05° precipitation climatology from the NASA Precipitation Measurement
Missions spaceborne radars: the TRMM Precipitation Radar (Ku-band, 1997–2014)
and the GPM Dual-frequency Precipitation Radar (DPR, Ku band, 2014–present).
Interactive map viewer:… See the full description on the dataset page: https://huggingface.co/datasets/snesbitt/pf-grid-tiles.osm-completeness-tiles219-tiles
Dataset Card
Dataset Description
[Placeholder]
Dataset Structure
[Placeholder]
Uses
[Placeholder]
Limitations
[Placeholder]
License
[Placeholder]
Citation
[Placeholder]
osm-saint-petersburg-5km-json-tilesTiling parameters:
lat0 = 59.946135
lon0 = 30.306768
tilesize_y=0.045
tilesize_x=0.08985372745626287
benchmarks-viz-tiles
Dexterous Manipulation Benchmark — Cross-Method Evaluation
Cross-method evaluation scaffolding for dexterous manipulation, comparing 4 methods × 4 hands × 2 datasets × 7 trajectories on identical hardware, scenes, and reference trajectories.
Tag: v0.7-mt-complete · Trained checkpoints in the private companion repo ckwolfe/benchmarks-trained-ckpts.
TL;DR
method
class
simulator
best metric
status
ManipTrans
closed-loop RL (residual)
IsaacGym
tracking err 4.0… See the full description on the dataset page: https://huggingface.co/datasets/ckwolfe/benchmarks-viz-tiles.instaroad-rosa-tilesSpatialData_with_spatial_tilesmulticolor_block_tilesThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so101",
"total_episodes": 181,
"total_frames": 61503,
"total_tasks": 26,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:181"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/vraiRobotLab/multicolor_block_tiles.urban-opengen-tiles-v1
Urban OpenGen tiles v1: the set the GAN was trained on
Sixty cities, the same 10 km window and 25 × 25 grid as
v2, rendered
from OpenStreetMap through a Grasshopper definition at 100 × 100 px and 4 m per pixel.
About 36,800 tiles present, of which 30,519 survive a 5% minimum-content filter.
You probably want v2 instead. It is finer at 3.125 m per pixel, its renderer is open
Python in the repository so you can extend it to any city yourself, and it carries the
whole-city rasters.… See the full description on the dataset page: https://huggingface.co/datasets/differential-studio/urban-opengen-tiles-v1.journeypulse-valhalla-tiles
JourneyPulse — Valhalla Offline-Kachelsatz Deutschland
Vorgebaute Valhalla-Routingkacheln für ganz
Deutschland, zur Nutzung im Offline-Routing der JourneyPulse-App.
Inhalt
valhalla_tiles.tar — unkomprimiertes Tile-Archiv (Valhalla liest es memory-mapped
als mjolnir.tile_extract), 1151 Kacheln
SHA256SUMS — Prüfsumme
Herkunft & Lizenz
Abgeleitet aus dem Deutschland-Gesamtextrakt von Geofabrik
(OpenStreetMap-Rohdaten). Die Daten stehen unter der
Open… See the full description on the dataset page: https://huggingface.co/datasets/HUHeine/journeypulse-valhalla-tiles.national-map-tiles
Vietnam road + building tiles, as PMTiles
Vector tiles (Mapbox Vector Tile) for the whole of Vietnam in PMTiles v3
archives, baked from an OpenStreetMap extract. This repository exists to be read over HTTP
range requests straight from a browser — Hugging Face serves 206 Partial Content with
Access-Control-Allow-Origin: *, which is what the PMTiles client needs and what most free
static hosts do not provide.
file
size
contents
national.pmtiles
430.4 MiB
roads… See the full description on the dataset page: https://huggingface.co/datasets/atemm99/national-map-tiles.tidetrace-sar-tiles
TideTrace SAR training tiles
512 x 512 tiles cut from the Zenodo Sentinel-1 SAR oil spill dataset, prepared
for the TideTrace oil spill attribution system (SIH26143, NTRO).
Contents
Each .npz holds one tile:
key
shape
meaning
image
(2, 512, 512) float32
Sigma0 VV and VH in dB
label
(512, 512) uint8
0 sea, 1 look-alike, 2 mineral oil
index.json records the tile count per class and the dB normalisation
statistics (mean_db, std_db) that the model… See the full description on the dataset page: https://huggingface.co/datasets/N-1ACE/tidetrace-sar-tiles.