geodot 0.1.14

Download satellite map tiles from the command line or as a Python library

geodot

geodot downloads satellite map tiles and prepares metadata-only visual place recognition (VPR) datasets. It can be used as a command-line tool or as a library from Python, JavaScript, and Rust.

Contents

• What It Does
• Install
• Quick Start
• CLI Reference
• Output Layout
• Dataset Preparation
• Dataset Contract
• APIs
• Tile Math
• Development

What It Does

Capability	Description
Tile download	Downloads 256x256 satellite tiles for a grid, rectangle, polygon, or GeoJSON area.
Demo page	Writes index.html for inspecting downloaded tiles over a MapLibre satellite basemap.
VPR preparation	Scans local imagery and writes virtual patch, variant, place, and quality manifests.
Validation	Checks prepared manifests, IDs, references, image paths, bboxes, and metadata-only flags.
Debug rendering	Renders one prepared single-tile patch or variant preview on demand.

geodot writes downloaded tiles directly from the tile server without re-compression.

Install

Choose the ecosystem that matches how you want to run or embed geodot.

Toolchain	Command	Provides
Rust	cargo install geodot	geodot CLI and Rust crate
npm CLI	npm install -g @geodot/cli	geodot CLI
npm library	npm install @geodot/lib	JavaScript library
Python	pip install geodot	geodot CLI and Python package

Supported runtimes are Node.js 18 or newer, Python 3.10 or newer, and Rust 1.85 or newer.

Run without a global npm or Python install:

npx -y @geodot/cli -x 37.6504907 -y 55.7303 -z 18 -c 1 -r 1
uvx geodot -x 37.6504907 -y 55.7303 -z 18 -c 1 -r 1

Rust uses cargo install geodot to install the CLI binary before running geodot.

Quick Start

Show CLI help or the installed package version:

geodot
geodot --help
geodot --version

Running geodot without arguments prints help and does not download tiles.

Download commands are explicit. For example, download a 3 by 3 grid of tiles at zoom 18:

geodot -x 37.6504907 -y 55.7303 -z 18 -c 3 -r 3 -o data -j 16

Serve the demo page:

geodot demo -o data

Then open http://127.0.0.1:8000/.

Prepare a metadata-only VPR dataset from downloaded tiles:

geodot --prepare -o data
geodot validate -o data

For a normal tile download, -o data writes:

data/
├── manifest.json
├── index.html
└── tiles/
    └── {z}/
        └── {x}/
            └── {y}.jpg

For dataset preparation, -o data writes:

data/
└── vpr/
    ├── manifest/
    │   ├── tiles.json
    │   ├── patches.json
    │   ├── variants.json
    │   ├── places.json
    │   └── quality.json
    └── config/
        └── dataset.json

CLI Reference

Selection Modes

Mode	Required arguments	Behavior
Grid	-x, -y, -z, -c, -r	Selects the tile containing the longitude/latitude and expands right/down by columns and rows.
Rectangle	-x, -y, --x2, --y2, -z	Selects all tiles between the two geographic corners. Argument order is top-left lon/lat and bottom-right lon/lat.
Polygon	-p, -z	Selects tiles whose center or corners are inside the polygon, plus tiles containing polygon vertices.
GeoJSON	--geojson, -z	Reads the first Polygon geometry from a Polygon, MultiPolygon, Feature, or FeatureCollection.

Polygon coordinates use lon,lat pairs. The closing edge is implicit.

Examples

Task	Command
Grid download	geodot -x 37.6504907 -y 55.7303 -z 18 -c 3 -r 3 -o data -j 16
Rectangle download	geodot -x 37.6504907 -y 55.7303 --x2 37.652 --y2 55.7297 -z 18 -o data
Polygon download	geodot -p "37.6504,55.7304;37.6520,55.7304;37.6520,55.7297;37.6504,55.7297" -z 18 -o data
Local GeoJSON download	geodot --geojson area.geojson -z 18 -o data
Remote GeoJSON download	geodot --geojson https://example.com/area.geojson -z 18 -o data
Remote country GeoJSON download	geodot -g "https://raw.githubusercontent.com/georgique/world-geojson/refs/heads/develop/countries/vatican.json" --out data
Prepare existing local tiles	geodot --prepare -o data
Download GeoJSON area and prepare	geodot --prepare --geojson https://example.com/area.geojson -z 18 -o data
Validate prepared dataset	geodot validate -o data
Validate with warnings as errors	geodot validate -o data --strict
Render a patch preview	geodot render -o data --patch-id <patch_id> --out preview.jpg
Render a variant preview	geodot render -o data --variant-id <variant_id> --out preview.jpg
Serve the demo	geodot demo -o data

Options

Flag	Default	Description
-h, --help	n/a	Print help and exit. Running geodot without arguments also prints help.
-v, --version	n/a	Print the installed package version and exit.
-x, --lon	none	Longitude for the grid/rectangle starting point. Required unless --polygon or --geojson is used.
-y, --lat	none	Latitude for the grid/rectangle starting point. Required unless --polygon or --geojson is used.
--x2, --bottom-right-lon	none	Rectangle bottom-right longitude.
--y2, --bottom-right-lat	none	Rectangle bottom-right latitude.
-p, --polygon	none	Polygon as lon,lat;lon,lat;lon,lat.
-g, --geojson	none	GeoJSON Polygon, MultiPolygon, Feature, or FeatureCollection file path or URL.
-z, --zoom	18	Web Mercator zoom level, 0 through 30.
-c, --cols	3	Tile columns to select in grid mode.
-r, --rows	3	Tile rows to select in grid mode.
-o, --out	data	Output directory.
-j, --jobs	16	Concurrent downloads.
--prepare	off	Prepare a VPR dataset. With selection arguments, downloads first and then prepares. Without selection arguments, scans an existing local dataset.
--patch-sizes	1,2,4,auto400m	Mosaic sizes in tiles for --prepare. Passing explicit values disables auto400m.
--stride	1	Tile stride for prepared mosaics.
--rotations	0,45,90,135,180,225,270,315	Rotation variants to record in metadata.
--no-manifest	off	Do not write manifest.json.
--no-demo	off	Do not write index.html.

Subcommands

Subcommand	Purpose	Notes
geodot demo -o data	Serves {out}/index.html over HTTP.	Do not open index.html with file://; browsers block local tile loading from file origins.
geodot validate -o data	Validates a prepared VPR dataset.	Exit code 0 means valid, 1 means validation errors, and 2 means missing dataset/manifests or invalid command usage.
geodot render -o data --patch-id <id> --out preview.jpg	Writes one debug preview for a patch.	Currently supports one-source-tile virtual patches.
geodot render -o data --variant-id <id> --out preview.jpg	Writes one debug preview for a variant.	Does not batch-render the dataset or compute descriptors.

validate checks that manifests parse, IDs are unique, references resolve, source images exist, bboxes and image dimensions are valid, dataset flags remain metadata-only, and preparation did not create images under vpr/. Warnings do not fail validation unless --strict is passed.

Tile URL Template

Set GEODOT_TILE_URL_TEMPLATE to use a different tile source:

GEODOT_TILE_URL_TEMPLATE='https://example.com/tiles/{z}/{x}/{y}.jpg' geodot -x 37.6504907 -y 55.7303 -z 18

The template can contain {sub}, {x}, {y}, and {z}. {sub} is one of mt0, mt1, mt2, or mt3.

Output Layout

For -o data, a 3 by 3 download at zoom 18 writes files like this:

data/
├── manifest.json
├── index.html
└── tiles/
    └── 18/
        └── 158488/
            └── 81979.jpg

manifest.json contains the selected center tile, downloaded tile records, and failed tile records:

{
  "center": { "x": 158488, "y": 81979, "z": 18 },
  "tiles": [
    {
      "tile": { "x": 158488, "y": 81979, "z": 18 },
      "bounds": {
        "lat_min": 55.730012,
        "lon_min": 37.650146,
        "lat_max": 55.730793,
        "lon_max": 37.651520
      },
      "path": "data/tiles/18/158488/81979.jpg",
      "bytes": 12345
    }
  ],
  "failed": []
}

The demo page reads tiles from {out}/tiles/{z}/{x}/{y}.jpg, overlays them on a satellite basemap, labels each tile as z/x/y, and keeps the downloaded imagery georeferenced while you zoom the view in or out. Large outputs are loaded lazily: the demo only mounts overlay images near the current viewport, so panning and zooming may update the displayed tile count in the panel. Use the transparency control to compare the downloaded overlay with the basemap, or open #z/x/y.jpg / /z/x/y URLs to center a specific tile.

Dataset Preparation

Preparation is metadata-only. It scans source images, validates coordinates, creates virtual patches/mosaics/variants, and writes manifests. It does not mutate source images, write cropped/rotated/resized/circular-masked/augmented images, compute descriptors, train models, or build ANN indexes.

Run preparation while downloading, after downloading tiles, or against any existing local folder that already follows the tile layout:

geodot --prepare --geojson https://example.com/area.geojson -z 18 -o data
geodot --prepare -o data
geodot --prepare -o data --patch-sizes 1,2,3 --stride 1 --rotations 0,90,180,270

geodot --prepare --geojson ... uses the normal GeoJSON tile download logic first, then immediately prepares the dataset. geodot --prepare -o data only scans an existing local dataset and does not download anything.

Source Image Roots

Preparation auto-detects these layouts:

Layout	Role	Capture ID
data/tiles/{z}/{x}/{y}.jpg	reference	default
data/tiles/{z}/{x}/{y}.jpeg	reference	default
data/tiles/{z}/{x}/{y}.png	reference	default
data/tiles/{z}/{x}/{y}.webp	reference	default
data/drone-view/{z}/{x}/{y}.jpg	query	default
data/drone-view/{z}/{x}/{y}.jpeg	query	default
data/drone-view/{z}/{x}/{y}.png	query	default
data/drone-view/{z}/{x}/{y}.webp	query	default
data/tiles/{capture_id}/{z}/{x}/{y}.jpg	reference	folder name
data/drone-view/{capture_id}/{z}/{x}/{y}.jpg	query	folder name

Supported source extensions are .jpg, .jpeg, .png, and .webp. Preparation reads image headers to record dimensions and detected format, but it does not require the extension to match the encoded bytes and never rewrites or converts source images.

drone-view/{z}/{x}/{y} assumes the drone image is already georeferenced to the corresponding Web Mercator tile location, for example by orthorectification or manual assignment to a known tile footprint. Zoom is a rough scale bucket, not UAV altitude. True altitude depends on camera FOV, sensor size, image resolution, pitch, terrain height, GSD, and whether the image is nadir or oblique. True UAV altitude and pose need optional metadata outside this filename convention.

Manual drone-view example using an externally supplied PNG:

mkdir -p data/drone-view/18/140140
curl -L "https://i.imgur.com/Aw7aFQb.png" -o data/drone-view/18/140140/97408.png
geodot --prepare -o data

Use .png for externally supplied PNG examples instead of saving PNG bytes under a .jpg filename.

Default Preparation Profile

Setting	Default
Native patches	1x1 tiles
Virtual mosaics	Overlapping 2x2 and 4x4 mosaics
Automatic scale	auto400m, clamped to 1-8 tiles
Zoom handling	All discovered zoom levels
Rotation variants	Metadata only
Circular crops	Metadata only
Appearance augmentation	None

auto400m estimates tile ground width at each patch latitude and zoom, chooses the nearest integer mosaic size for roughly 400 meters, clamps it to 1-8 tiles, and deduplicates it with explicit patch sizes. At zoom 18 near the default latitude this is usually around a 5x5 patch, matching the scale of LASED-like aerial VPR datasets.

Prepared Files

File	Contents
vpr/manifest/tiles.json	One record per discovered source tile with tile ID, root, capture ID, role, path, z/x/y, image size, byte size, bbox, center lon/lat, and validity.
vpr/manifest/patches.json	One record per native tile or complete mosaic window with place ID, source tile IDs, pixel size, bbox, center lon/lat, estimated ground size, circular-crop availability, and a virtual compose spec.
vpr/manifest/variants.json	Virtual rotations with empty descriptor/index IDs for downstream descriptor extraction.
vpr/manifest/places.json	Matching reference/query imagery grouped by z/x/y location.
vpr/manifest/quality.json	Conservative, non-destructive low-information labels from cheap image/file statistics.
vpr/config/dataset.json	Dataset profile, schema version, detected roots, zoom levels, preparation settings, and counts.

Mosaics, rotations, and circular crops are virtual by default. A patch points to source tile IDs and layout instructions instead of writing new JPEGs, keeping storage small and source tiles immutable.

Dataset Contract

geodot creates image datasets and virtual patch metadata only. It intentionally does not extract descriptors, train models, build ANN indexes, or choose model-specific preprocessing.

External descriptor tools should treat these IDs as stable dataset IDs:

ID	Meaning
tile_id	Source image tile identity.
place_id	Geographic z/x/y place identity.
patch_id	Native tile or virtual mosaic patch identity.
variant_id	Patch plus virtual transform identity.

The usual downstream Python flow is:

from geodot import load_dataset, render_variant

dataset = load_dataset("data")
image = render_variant(dataset, variant_id)
# pass image bytes to DINOv3 / ResNet / VLAD / GeM outside geodot

Descriptor outputs should live outside geodot, or in a separate downstream folder, and reference variant_id. Re-running descriptors with DINOv3 SAT, ResNet, VLAD, GeM, NetVLAD, steerable CNNs, or another model should not require changing source images or regenerating the prepared dataset.

APIs

Python API

from geodot import (
    Coordinate,
    DownloadOptions,
    PrepareOptions,
    download,
    latlon_to_tile,
    prepare_dataset,
    tile_bounds,
    tile_grid,
    tile_grid_between,
    tile_grid_for_polygon,
)

tile = latlon_to_tile(55.7303, 37.6504907, 18)
bounds = tile_bounds(tile)
tiles = tile_grid(55.7303, 37.6504907, zoom=18, cols=3, rows=3)
rectangle = tile_grid_between(55.7303, 37.6504907, 55.7297, 37.652, zoom=18)
polygon = tile_grid_for_polygon([
    Coordinate(lon=37.6504, lat=55.7304),
    Coordinate(lon=37.6520, lat=55.7304),
    Coordinate(lon=37.6520, lat=55.7297),
    Coordinate(lon=37.6504, lat=55.7297),
], zoom=18)

report = download(DownloadOptions(
    lat=55.7303,
    lon=37.6504907,
    bottom_right_lat=55.7297,
    bottom_right_lon=37.652,
    zoom=18,
    cols=3,
    rows=3,
    out="data",
    jobs=16,
    geojson=None,
    no_manifest=False,
    no_demo=False,
))

dataset = prepare_dataset(PrepareOptions(out="data"))

JavaScript API

import {
  download,
  latlonToTile,
  prepareDataset,
  tileBounds,
  tileGrid,
  tileGridBetween,
  tileGridForPolygon,
} from '@geodot/lib';

const tile = latlonToTile(55.7303, 37.6504907, 18);
const bounds = tileBounds(tile);
const tiles = tileGrid(55.7303, 37.6504907, 18, 3, 3);
const rectangle = tileGridBetween(55.7303, 37.6504907, 55.7297, 37.652, 18);
const polygon = tileGridForPolygon([
  { lon: 37.6504, lat: 55.7304 },
  { lon: 37.6520, lat: 55.7304 },
  { lon: 37.6520, lat: 55.7297 },
  { lon: 37.6504, lat: 55.7297 },
], 18);

const report = await download({
  lat: 55.7303,
  lon: 37.6504907,
  bottomRightLat: 55.7297,
  bottomRightLon: 37.652,
  zoom: 18,
  cols: 3,
  rows: 3,
  out: 'data',
  jobs: 16,
  geojson: undefined,
  noManifest: false,
  noDemo: false,
});

const dataset = await prepareDataset({ out: 'data' });

Rust API

use geodot::{
    Coordinate, DownloadOptions, PrepareOptions, download, latlon_to_tile, prepare_dataset,
    tile_bounds, tile_grid, tile_grid_between, tile_grid_for_polygon,
};

#[tokio::main]
async fn main() -> anyhow::Result<()> {
    let tile = latlon_to_tile(55.7303, 37.6504907, 18);
    let bounds = tile_bounds(tile);
    let tiles = tile_grid(55.7303, 37.6504907, 18, 3, 3);
    let rectangle = tile_grid_between(55.7303, 37.6504907, 55.7297, 37.652, 18);
    let polygon = tile_grid_for_polygon(&[
        Coordinate { lon: 37.6504, lat: 55.7304 },
        Coordinate { lon: 37.6520, lat: 55.7304 },
        Coordinate { lon: 37.6520, lat: 55.7297 },
        Coordinate { lon: 37.6504, lat: 55.7297 },
    ], 18);

    let report = download(DownloadOptions {
        lat: 55.7303,
        lon: 37.6504907,
        bottom_right_lat: Some(55.7297),
        bottom_right_lon: Some(37.652),
        polygon: Vec::new(),
        geojson: None,
        zoom: 18,
        cols: 3,
        rows: 3,
        out: "data".into(),
        jobs: 16,
        tile_url_template: None,
        no_manifest: false,
        no_demo: false,
    })
    .await?;

    let dataset = prepare_dataset(PrepareOptions {
        out: "data".into(),
        patch_sizes: vec![1, 2, 4],
        stride: 1,
        rotations: vec![0, 45, 90, 135, 180, 225, 270, 315],
        auto400m: true,
    })?;

    Ok(())
}

Tile Math

Given tile { z: 18, x: 158488, y: 81979 }:

n = 2^z
lon_min = x / n * 360 - 180
lon_max = (x + 1) / n * 360 - 180
lat_max = atan(sinh(pi * (1 - 2y/n))) * 180/pi
lat_min = atan(sinh(pi * (1 - 2(y+1)/n))) * 180/pi

Tile bounds are represented as lat_min, lon_min, lat_max, and lon_max fields.

Approximate resolution near the default latitude (55.7303):

Zoom	m/px	Tile covers
18	0.34	86 x 86 m
16	1.36	347 x 347 m
14	5.45	1.4 x 1.4 km
10	86	22 x 22 km
8	345	88 x 88 km

Development

Install local Python test dependencies and run the Python tests:

python -m pip install -e '.[test,dev]'
python -m pytest

Run JavaScript and Rust tests:

npm install
npm test
cargo test --manifest-path rust/Cargo.toml

Run lint and format checks:

python -m pip install -e '.[test,dev]'
ruff format --check python
ruff check python
npm run format:check
npm run lint
cargo fmt --manifest-path rust/Cargo.toml -- --check
cargo clippy --manifest-path rust/Cargo.toml --all-targets -- -D warnings

Install the local pre-commit hook to run format, lint, and tests before every commit:

cp scripts/pre-commit .git/hooks/pre-commit
chmod +x .git/hooks/pre-commit