Rail-mounted camera
Linear gantry · auto-pan
Inspection robot on a rail. Scans rows on a loop. RTSP / MQTT pushes frames to UPII.
Drone, pole, phone, spectrometer — same Neural Spine.
Four real deployments — open field, fruit cluster, elevated rows, polytunnel. Scroll to walk through. Each stage carries one detection and one action the substrate took.
YOLO + vision LLM cross-validate every detection. Field manager gets a triage list before lunch.
A sensor that only a human can read is visibility. A sensor whose output the agent can act on is legibility. Every camera, every phone, every spectrometer enters through the same ingress.
Solid bbox = confidence at threshold. Dashed = below threshold, surfaced for human re-capture. Never silent.
Live deployment runs on the substrate today. Production access is by engagement — talk to us.
UPII accepts every protocol below. Schema negotiation on first contact. Binary-safe transport. Per-device crypto identity. The machine you already own is the machine the substrate uses.
Linear gantry · auto-pan
Inspection robot on a rail. Scans rows on a loop. RTSP / MQTT pushes frames to UPII.
Tripod · 24fps
Wide-angle on a static post. UPII signs each frame, drops dupes by replay nonce.
iOS Safari · Android Chrome
Phone in the field. Under-10-second diagnosis + agronomic advice. Feedback re-trains the model.
Hamamatsu C15511 · USB-HID
Lab-grade spectral signature. Late-fusion with YOLO bbox. Material / sugar-content readouts.
6-DOF · soft gripper
Autonomous picking + selective spray. Triggered by Workflow on YOLO confidence ≥ threshold.
Strawberries today. Coffee, tea, leafy greens — same substrate, different crop. Different chains: physical anomaly → twin state → workflow → audit log.