OPERATION XRAY PROTOTYPE / 01
INTELLIGENCE AT THE CHECKPOINT

A second look.
Before the gate.

A copilot for baggage X-ray operators. Bring regions worth reviewing into focus, and keep every screening decision with the human.

Enter the operator demo Explore the concept

Synthetic demo · No validated threat detection

Concept rendering of a baggage X-ray scanner with an operator review station
VENUE BAGGAGE EDGE REVIEW CONCEPT
03Region surfacedOperator review required
Designed around the operator.Baggage-first • Human-led • Audit-aware
01 / THE STARTING POINT

One focused workflow.
A clear path to evidence.

Crowded entrances deserve careful review. Our starting hypothesis: a second set of eyes can help operators find subtle or overlapping objects in baggage imagery.

This prototype makes that workflow tangible. A small, learned model examines synthetic silhouettes; the console shows the region, the model's limits, and the operator's recorded decision.

Concert venuesSports arenasBaggage screening
02 / THE PRODUCT

Assist attention.
Preserve judgment.

Try all six cases
01

Receive a scan

Concept integration with an existing baggage scanner. The working demo uses only procedural images.

02
03

Surface a region

Inspect the silhouette, toggle the overlay, and adjust zoom or contrast. A score is a cue for attention.

03
REVIEWED↗

The operator decides

Record review complete, escalation, or a benign disposition. Export the local session audit.

03 / THE SYSTEM CONCEPT

A software layer.
Beside the hardware.

Explore an operator workstation and edge inference module alongside existing scanners. The enclosure and connection paths are concepts for a future integration study.

  • 01 Imaging stays in the screening workflow
  • 02 Model cues remain explainable and reviewable
  • 03 Operator dispositions create a feedback loop
See the working review loop
PROPOSED INTEGRATION / NOT CONNECTED
▥
Existing scannerImage acquisition
01
↓ Vendor-approved image interface
✳
Assistance moduleLocal inference · region cues
02
↓ Reviewable findings
▧
Operator consoleHuman decision · local audit
03
This demo runs locally in your browser.
04 / WHAT COMES NEXT

Build the evidence
before the claim.

A staged pilot proposal, with explicit gates between a concept and operational use.

NOW

Demonstrate the human review loop

Interactive console, synthetic inference, CAD concept, and a decision export.

01
NEXT

Validate with operators and scanner partners

Confirm workflow fit, data access, and an approved path to representative imagery.

02
GATE

Evaluate under controlled conditions

Measure missed objects, false alerts, review burden, and operator performance before any live screening claim.

03
↗
Start with venue baggage. Explore other markets later.

Aviation security and medical imaging need separate datasets, workflows, and validation. MRI is a distinct imaging modality; this demo performs no medical diagnosis or imaging of people.

05 / TAKE IT WITH YOU

The pitch. The prototype.
The hardware concept.

Review the opportunity and explore the tangible concept behind the workflow.

PDF
Investor pitchPresentation for review
PPTX
Editable deckPowerPoint source
3D
Hardware sceneNative Blender concept
GLB
Edge appliancePortable 3D model
THE WORKING PROTOTYPE

See what the operator sees.

Six synthetic bags. A learned shape model. A human decision.

Start a review session
OPERATION XRAYConcept prototype · October 2026
TRANSPARENCY / MODEL CARD

A real toy model.
A clear limit.

A logistic classifier runs in your browser on procedurally generated object masks. It is trained to distinguish one tapered, guarded blade-like silhouette family from everyday synthetic shape families.

MODELLogistic regression
INPUT32 × 32 binary masks
FEATURES9 silhouette measures
DATASynthetic shapes only

What is actually working

Pixel-derived shape features are computed at runtime and passed through exported learned coefficients. Case IDs, case names, and object labels are not inputs to inference. Zoom, contrast, region selection, operator recording, session restart, and JSON export work locally.

Held-out synthetic evaluation

These figures describe only same-generator, held-out masks. They say nothing about real X-ray detection, safety, medical diagnosis, operator effectiveness, or regulatory acceptance.

Known limitations

Before a real pilot

Obtain an approved scanner interface and representative labeled imagery, define the prohibited-object policy with operators, train a model suited to the data, and measure false alerts, missed objects, and review burden in a controlled human-factors study.

Reproducible seed: 731 · model: silhouette-logistic-v1Download model coefficients ↗
BROWSER-LOCAL / APPEND-ONLY DURING SESSION

Operator audit

Each record captures the viewed region, learned shape score, disposition, and optional note. Restarting the demo begins a new run and preserves previous records.

Stored on this device when browser storage is available. This prototype is editable through browser tools and is not a compliance ledger.
NEW DEMONSTRATION RUN

Restart the queue?

All six synthetic bags return to the queue. Previous decisions stay in the audit so you can compare runs.