Goatie
Computer vision · Re-identification · Edge and cloud hybridGoatie is a computer vision and records platform that evaluates animal movement, structure and handling from ordinary phone video and maintains a longitudinal record across a season. It is deployed for youth agricultural programs.
Object detection and animal re-identification across sessions on a hybrid edge and cloud architecture, with explainable outputs that carry reasoning and a confidence level. Applicable to any monitoring task requiring identity persistence across captures.
Detection and re-identification run at the edge with cloud aggregation. Supporting models estimate mass from body measurements, extract structured data from receipts, and surface trends across longitudinal records. Every inference states its reasoning and confidence, and the system is positioned as decision support.
Detection and re-identification at 6.5 ms per frame, AP50 0.88, on the hybrid architecture.