gtfs2geojson 0.5.0

Convert GTFS feeds to GeoJSON with optional Folium preview maps.

gtfs2geojson

Convert any GTFS feed (zip file or directory) into a clean, styled GeoJSON FeatureCollection, with an optional interactive Folium preview map.

Built for transit data folks who want a quick, dependency-light way to go from gtfs.zip → something they can drop into kepler.gl, QGIS, deck.gl, Mapbox, or a Leaflet/Folium map — without spinning up Postgres/PostGIS.

Features

• Single command, no database: gtfs2geojson feed.zip -o feed.geojson
• Reads zip or directory GTFS sources, handles UTF-8 BOM
• Routes as MultiLineString features, one per route_id (all shape variants merged)
• Stops as Point features
• Mode-aware styling — readable labels (Bus / Metro / Tram / Ferry / Trolleybus / …) with sensible default colours, falling back from route_color when missing
• Reconstructs missing geometry from stop_times when a route has no entry in shapes.txt (common for metros/trams in older feeds)
• Filters: by mode (--mode Bus), by agency (--agency OASA), by bounding box (--bbox w,s,e,n), by service day (--date 20260615)
• Line simplification with Ramer-Douglas-Peucker (--simplify 0.0001) for smaller, faster-rendering output
• Output formats: standard GeoJSON or GeoJSON-seq (--format geojsonseq) for tippecanoe / ogr2ogr pipelines
• Introspection: --list-modes and --list-agencies (with --json) to peek at a feed without converting
• Optional Folium preview with per-mode toggleable layers, mode legend, summary sidebar, stop search, and configurable tiles (--tiles dark_matter)
• Extended GTFS route types supported (HVT codes 100–1700)

Install

pip install gtfs2geojson           # core
pip install "gtfs2geojson[preview]" # also installs folium for --preview

Or from source:

git clone https://github.com/acesonic7/gtfs2geojson
cd gtfs2geojson
pip install -e ".[preview]"

CLI

# basic
gtfs2geojson feed.zip -o feed.geojson

# with interactive map (CARTO Positron tiles by default)
gtfs2geojson feed.zip -o feed.geojson --preview map.html

# dark basemap for nighttime / dataviz
gtfs2geojson feed.zip --preview map.html --tiles dark_matter

# filter to buses + metro inside a bbox, skip stops
gtfs2geojson feed.zip \
  --mode Bus --mode Metro \
  --bbox 23.6,37.9,23.8,38.05 \
  --no-stops \
  -o athens_core.geojson

# only trips active on a particular calendar date (combines calendar.txt
# weekday rules with calendar_dates.txt exceptions)
gtfs2geojson feed.zip --date 20260615 -o that_monday.geojson

# simplify polylines for web/mobile rendering (~10 m at most latitudes)
gtfs2geojson feed.zip --simplify 0.0001 -o feed.simplified.geojson

# pipe straight into tippecanoe (GeoJSON-seq, one Feature per record)
gtfs2geojson feed.zip --format geojsonseq | tippecanoe -o feed.mbtiles

# peek at what a feed contains, without converting
gtfs2geojson feed.zip --list-modes
gtfs2geojson feed.zip --list-agencies --json | jq '.[] | select(.n_routes > 5)'

# feed already extracted to a directory
gtfs2geojson ./gtfs_extracted/ -o feed.geojson

Stream to stdout (default) for piping into other tools:

gtfs2geojson feed.zip | jq '.features | length'

Python API

from gtfs2geojson import convert, write, list_modes, list_agencies
from gtfs2geojson.preview import render

geo = convert(
    "feed.zip",
    modes=["Bus", "Trolleybus"],
    bbox=(23.6, 37.9, 23.8, 38.05),
    service_date="2026-06-15",       # also accepts "20260615" or datetime.date
    simplify_tolerance=0.0001,       # ~10 m perpendicular tolerance
    include_stops=True,
)
write(geo, "feed.geojson")                    # default
write(geo, "feed.geojsonseq", format="geojsonseq")  # one Feature per RFC-8142 record
render(geo, "preview.html", title="Athens core network", tiles="dark_matter")

# Peek at a feed without converting
list_modes("feed.zip")        # -> {"Bus": 42, "Metro": 3, ...}
list_agencies("feed.zip")     # -> [{"agency_id": "OASA", "agency_name": "OASA", "n_routes": 45}, ...]

Output schema

Each route feature carries:

Property	Description
feature_type	always "route"
route_id, route_short_name, route_long_name	from routes.txt
route_type	original GTFS integer
mode	human label (Bus, Metro, Tram, Ferry, …)
route_color, route_text_color	hex strings, with mode-based defaults
agency_id, agency_name	joined from agency.txt
headsigns	up to 6 distinct headsigns concatenated with |
shape_source	"shapes.txt" or "stop_sequence" (reconstructed)
n_trips	number of trips serving this route
n_stops	number of distinct stops touched by any trip on this route
length_km	total network length, sum of all shape variants (haversine, km)

Stop features carry:

Property	Description
feature_type	always "stop"
stop_id, stop_name, stop_code	from stops.txt
route_ids	sorted list of routes serving this stop
modes	sorted list of distinct modes among those routes

By default, stops not served by any kept trip are dropped — pass --keep-orphan-stops (CLI) or keep_orphan_stops=True (Python) to retain them. If stop_times.txt is absent, all stops are kept regardless.

Why?

Converting GTFS to GeoJSON should be a one-liner. Most existing options either:

• require Postgres/PostGIS (too heavy for a quick visualisation),
• assume shapes.txt is always present (it isn't — many metros, trams, and older feeds skip it),
• or output undecorated geometries with no styling hooks.

gtfs2geojson aims to be the boring, dependable, do-one-thing tool you can plug into a Makefile or notebook.

Limitations / non-goals

• No frequency or schedule data on the output.
• No realtime (GTFS-RT) support — schedule feeds only.

If you need any of the above, gtfs_kit and partridge are excellent.

License

MIT.