DIU CSO Endpoint Accuracy 2 — Part 1

Aided Target Recognition for CROWS, demonstrated.

A non-intrusive AiTR overlay. Detection, multi-object tracking, monocular ranging, and kinematic threat assessment on commodity edge compute — with a mandatory human-in-the-loop gate and graceful fallback to standard RWS operation if the pipeline degrades.

Try the live demo Read the brief

Live operator console

Below is the same pipeline submitted with the DIU application. Switch the perturbation profile to Engagement or Severe to see jitter and muzzle-flash injection. Click Disable AiTR to exercise the failsafe.

Demo backend not yet wired

The CF Pages site is live, but the backend URL still needs to be set in app.js once the Python server is hosted (Fly.io / HF Space / VPS). In the meantime:

Recorded walkthrough

3–5 minute screen capture of the system. Used when the live backend is asleep, and the canonical reference for reviewers.

Recorded walkthrough — link coming soon

What the demo shows

Live AiTR loop

Threat-colored boxes on tracked targets with class, range, confidence, and HITL action buttons (Confirm / Engage / Dismiss).

Adverse-conditions toggle

Inject jitter, muzzle flash, vibration, and dropout on demand. Directly addresses the CSO clause on "shake from weapon firing, muzzle flash, and high frequency jitter."

HITL hard gate

Every engagement click triggers a confirmation modal and is recorded in the audit log. The system never issues a slew command without an operator action.

Failsafe

FPS or detector errors auto-degrade the pipeline; the operator can also disable AiTR with one click. Raw video continues; the operator reverts to manual RWS control.

What the artifact is