Strava Globe

Your entire Strava history as glowing tracks on a realistic 3D Earth.

Point it at the export zip Strava gives you, and you get a private, local globe (CesiumJS + satellite imagery) you can spin, zoom from orbit down to street level, and fly around:

• Every GPS track drawn on the globe, colored by activity type, with hover tooltips (name · date · distance · place) — click a track to fly to it.
• ▶ Tour mode: your activities are clustered into places, named via offline reverse geocoding, ordered by first visit — and the camera flies through your travel history stop by stop, with a card per place and a closing summary.
• Your photos, where you took them: pictures attached to activities show up on the activity card and on tour stops — click one for a full-size lightbox. App-generated stat cards (training-plan summaries and the like) are detected and skipped, so only real photographs appear.
• Filters by activity type and year, satellite or street basemap, and an idle globe spin until you grab it.
• No accounts, no API keys, no upload. Your tracks never leave your machine.

Tour mode mid-journey: Rawai → Ao Nang, each stop card carrying the photos taken there — click one for full size.

From orbit to city…	…to street level

Quickstart

1. Request your archive on strava.com: Settings → My Account → Download or Delete Your Account → Request Your Archive (the zip arrives by email).

2. Run (with uv, no install needed):

   uvx strava-globe ~/Downloads/export_12345678.zip
   

   or classically:

   pip install strava-globe
   strava-globe ~/Downloads/export_12345678.zip
   

The globe builds and opens at http://localhost:8933. Next time, plain uvx strava-globe reuses the last build (the generated data lives in your user data directory; --port, --no-browser, and --data-dir are available).

How it works

The builder (strava_globe/tracks.py) reads activities.csv from the export — zip or extracted folder — then for every activity file:

• decodes FIT with fitfast and GPX with a small XML parser (indoor activities without GPS are skipped),
• splits tracks at GPS dropouts (>500 m between fixes),
• simplifies each segment with Ramer-Douglas-Peucker (~5 m tolerance, simplification),
• tags the activity with its nearest city and country — offline GeoNames lookup via reverse-geocoder,
• thumbnails any photos attached to the activity (EXIF-rotated, two sizes, Pillow) into your local data directory — dropping app screenshots, detected when an image is both locally flat (synthetic gradients, share of near-zero variance 8×8 blocks) and extremely compressible (bytes per pixel); real photographs carry sensor noise everywhere and never trip both,
• and writes one compact activities.json (~1.7 MB for 725 tracks / 1.9 M raw GPS points).

The web app (strava_globe/web/) is dependency-free: CesiumJS 1.131 from the jsdelivr CDN, imagery streamed from Esri World Imagery (satellite) or OpenStreetMap (streets), served by a stdlib static server with caching disabled.

Controls

Privacy

Your GPS data stays local: the generated activities.json lives in your user data directory (never in this repo or the package), and the app talks only to the CDN and the map tile servers (which necessarily see which map areas you view, as with any online map).

Notes

• Rendering quirk: msaaSamples = 1 and dynamicAtmosphereLighting = false are deliberate — MSAA and sun-driven atmosphere shading render a black globe on software-GL stacks (e.g. headless/embedded Chromium).
• Tile imagery: Esri World Imagery is used under its terms for personal/non-commercial viewing; OpenStreetMap tiles per the OSM tile usage policy.
• Not affiliated with or endorsed by Strava, Garmin, Esri, or OpenStreetMap. "Strava" is used descriptively: the input is the data export Strava provides to its users.

Development

git clone https://github.com/lukaszkolodziejczyk/strava-globe
cd strava-globe
uv run strava-globe path/to/export.zip

Releases: push a v* tag — GitHub Actions builds and publishes to PyPI via trusted publishing (.github/workflows/release.yml).

License

BSD 3-Clause. Place names derived from GeoNames data (CC-BY 4.0) via reverse-geocoder.