Book a call
All use cases

01 · Queryable training data

85,000+ images.
Only the right ones matter.

AI Library turns raw field imagery into searchable training data. Every image stays connected to its field, crop, growth stage, weed, device and capture date — making it possible to isolate exactly the conditions a model needs.

Build a training subset
  1. CropSpinach
  2. Growth stage11 weeks
  3. WeedBrandnetel
  4. DeviceRobotOne

242matching imagesfor this training condition

AI Library map with crop, growth-stage, weed and device filters applied

02 · Spatial provenance

Every image stays connected to the field it came from.

Images remain linked to the field observation that produced them — including capture date, crop conditions and device metadata. The dataset keeps its real-world origin visible all the way down to an individual photo.

AI Library interface showing an observation, field-image grid and its selected field on the map
Traceable by design One record. Three connected layers.
  1. FieldBloemkool, zomer, productie

    Spatial origin of the dataset

  2. ObservationPixelFarming Import · Jul 11, 2026

    Capture event with agricultural metadata

  3. Imagery1256 field photos

    Images stay linked to their source context

03 · Spatial architecture

GIS scale without
GIS friction.

AI Library uses vector tiles to stream only the geographic detail needed for the current view — keeping the map fast from country scale down to individual fields.

  1. 01 · Data

    PostgreSQL
    + PostGIS

    Field boundaries, observations and image metadata live in one spatial data layer.

  2. 02 · Delivery

    Vector
    tiles

    Only the geometry needed for the current viewport and zoom is sent to the browser.

  3. 03 · Interface

    MapLibre
    + React

    Users move smoothly from regional overview to individual fields and observations.

04 · User-owned data

Your fields.
Your imagery.
Your data.

Ownership and access are part of the data model, not an afterthought. Rights stay attached to field imagery from the moment it enters the library.

All fieldsMy fields
MineCreated and controlled by me
OwnedShared with access rights
UnownedDiscoverable in the wider dataset
FreePaid

Access terms travel with the observation.

Explore globally.Manage locally.

Portable identity and user-controlled storage keep the platform from becoming another data silo.

05 · One stack, one team

From spatial schema
to production interface.

One team designed the domain model, map delivery, application layer, user experience and release coverage as a connected product.

01

Spatial browsing

React · MapLibre

A responsive map-first interface for exploring fields and observations.

02

Application layer

Node.js · Hono · tRPC

A typed path from the interface to domain data and actions.

03

Geometry + metadata

PostgreSQL · PostGIS · Drizzle

Field boundaries, observations and image context in one model.

04

Identity + ownership

W3ID · W3DSeVault

Portable identity and a user-controlled data layer.

05

Release confidence

Playwright

Critical product flows checked before every deploy.

Have a geospatial product in mind?

Let’s turn domain data
into a product people use.

Discuss your project