Case Study · Apple Yield Detection

End-to-end computer-vision example from the NaLamKI validation phase.

Goal

Detect apples on tree imagery and estimate yield at the orchard level.

Architecture

A stateless CV service consumes high-resolution images from the Spatio-Temporal API (drone or hand-held captures stored as items in a STAC collection), runs detection, and writes the count + bounding boxes back as vector features in a feature collection on the same API — directly visualisable via MVT vector tiles.

APIs used

• Spatio-Temporal API — input imagery via STAC; output detections via POST /api/v2/collections/{id}/items and the WebMercatorQuad-Vector tileset
• Digital Farm API — orchard / row / tree regions
• Service Registry — commissioning, results

Pattern

This is the Computer Vision pattern in Build · AI Services. Use it as a template for any image-in / detections-out service.