About Services Tools Blog Contact us

Open-Map Place Clustering

Builds a country's campsites, shelters, viewpoints, beaches and food stops from free map data, scores them by type of trip, and groups them into named weekend hotspots.

  • Mapping & Geospatial
  • Leisure and sport
  • Pipeline
  • Product
  • Hybrid
  • 2026

The problem

A destination map stays empty until someone turns raw map data into named, scored groups a visitor can browse. Entering every campsite and viewpoint by hand would take far too long. The raw data is also too noisy to search directly.

What it does

You give the tool a country code. It downloads that country's OpenStreetMap data from Geofabrik and turns it into categorised places, dropping shapes that are only part of a larger object. Trip-type scores come from simple rules, or from the LFM2 model running through Ollama. Nearby places are then grouped into named hotspots and indexed, so a phrase such as a quiet coastal campsite can be searched.

Outcomes

  • A country of places appears with no hand entry.
  • Hotspots are weekend-sized groups and the map does not show every pin at once.
  • Rule-based scoring still runs when no model is available.
  • A phrase search can find a style of place as well as a name.

Simplified mockup

Interactive preview with sample data. Nothing is sent to a server.

Cluster a region

Region cached extract

Relation-only geometry is dropped before scoring. Trip style uses heuristics while no language model is available.

Stay Activity Food 5 km join
Pipeline
Ready
  1. Extract
  2. Drop geometry
  3. Label
  4. Score
  5. Cluster
  6. Name
Places in extract
1,842
After the tag filter
Relation members dropped
316
Kept as part of a parent
Scoring
Rules
Model offline

Cluster membership

Hotspots
0
Linked places
0
Join radius
5 km
Sample members
PlaceCategoryHotspotFrom centre
Reed Bed CampCampingSaltmere Shore0.4 km
Harbour MeadowCampingSaltmere Shore1.1 km
Shore BothyShelterSaltmere Shore0.8 km
Salt Kiln KitchenFoodQuay Kitchens0.3 km
Net Loft SupperFoodQuay Kitchens0.6 km
Westcombe BeaconViewpointWestcombe Head0.2 km
Highfen BothyShelterHighfen Bothies1.4 km
Pebble Reach SwimBeachPebble Reach0.5 km

A place stays linked while it sits inside the hotspot radius. Pins outside every group stay on the country map as single sites.

Search by phrase

Northstrand clusters
18Saltmere Shore
16Pebble Reach
11Quay Kitchens
14Westcombe Head
9Highfen Bothies
5 km
Highlighted cluster matches the phrase
Ranked clusters
ClusterWhyScore
Saltmere Shore best Quiet stay on the coast 0.91
Pebble ReachBeach weekend, busier shore0.62
Quay KitchensFood stops, not a pitch0.33
Westcombe HeadViewpoint circuit0.28
Highfen BothiesInland shelter0.21

How it works

  1. Download the map data for a country code and filter it into places.
  2. Drop shapes that are only part of a larger object.
  3. Turn tags and names into stay, activity and food categories, plus a region.
  4. Write trip-type scores by simple rules, or with a model when one is available.
  5. Group nearby places into hotspots with a centre, an outline and a list of members.
  6. Name and index the groups so the map can search by phrase as well as by area.

Pipeline

flowchart LR
  Code[Country code] --> Extract[Filtered extract]
  Extract --> Label[Categories]
  Label --> Score[Trip-style scores]
  Score --> Cluster[Distance clusters]
  Cluster --> Name[Names and membership]
  Name --> Site[Map can browse them]