Stud Striker AI · FAQ
Frequently asked questions
Answers for property operators, facility managers, and technical teams evaluating or running Stud Striker AI at a garage or lot entrance.
General
01What does Stud Striker AI actually do?
It watches your entry lane with a camera and automatically tells you whether an approaching vehicle is running studded tires or not — in under a second, with no one standing at the gate. Every decision is logged, timestamped, and tied to a specific lane and property.
Why would a property need this?
Studded tires damage pavement and are restricted by season in many jurisdictions. Today, enforcing that means a person physically checking tires — which doesn’t scale, isn’t consistent, and leaves no record. Stud Striker turns that into an automated, continuous, defensible process.
What counts as a “decision”?
Every frame the system evaluates resolves to one of three states: studded, non-studded, or undetermined. Undetermined means the model’s confidence was below your configured threshold — the system doesn’t guess, it just keeps watching for a clearer frame.
Does it stop or block vehicles?
It can. Stud Striker integrates with garage entry systems: on a studded-tire detection it can display feedback directly on the entry system’s own hardware (screens/signage at the lane) and control the swing arm — holding or denying entry per your property’s policy. Detection and enforcement are configured separately, so you can also run it in detect-and-log-only mode.
Which entry systems do you integrate with?
We integrate with garage entry systems to drive their existing lane hardware — on-lane displays and swing-arm control — rather than requiring separate gates or signage. Compatibility is confirmed per site during the pilot; tell us what system your property runs and we’ll verify the integration path.
Accuracy & Detection
02How accurate is the detection?
Detection runs on a YOLOv8 model trained specifically on stud patterns, scored against a confidence threshold (default 0.75, tunable per lane). Frames below threshold are marked undetermined rather than forced into a wrong answer — the system is tuned to avoid false positives, not to force a call on every frame.
What happens in bad lighting, snow, or at night?
Confidence drops in poor conditions, which increases the undetermined rate rather than producing bad calls. Camera placement and lighting at the lane matter — we advise on positioning during setup to keep the undetermined rate low.
Can it tell studded from worn-down winter tires?
Yes — that distinction is exactly what the model is trained on. Worn or missing studs, standard all-seasons, and true studded tread are visually distinct enough for the model to separate reliably at the trained confidence threshold.
Can the model be retrained or improved over time?
Yes. Swapping in a retrained model file requires no application code changes — new detection classes or improved accuracy ship as a model update, not a system rebuild.
Hardware & Setup
03What hardware do we need at each lane?
A camera pointed at the approach — Arducam Pivistation, Raspberry Pi Camera, an existing IP/streaming camera, or a USB webcam all work. If your property already has camera infrastructure at the entrance, we can often point Stud Striker at the existing feed rather than installing new hardware.
How long does install take per lane?
A single-lane pilot is typically a same-day setup once the camera is mounted and network-reachable — configuration is a single file naming the lane, building, and confidence threshold.
Does each lane need its own installation?
Each lane runs its own camera and configuration entry, but scaling to more lanes reuses the same software — you’re adding a config block and a camera, not standing up a new system.
What if our camera hardware is overtaxed or unreliable?
Edge devices like the Raspberry Pi can be sensitive to sustained load. Non-essential services (like verbose logging) can be disabled per-lane, and running detection locally rather than over a remote connection avoids the most common stability issues.
Data, Privacy & Audit
04What gets logged, and where does it go?
Structured, timestamped detection logs go to Grafana Loki for centralized search. System and detection metrics (confidence, throughput, resource use) are exported to Prometheus for live dashboards. Frame images are archived to Google Drive — selectively, only when a decision required review — to keep storage lean.
Is this a surveillance system? Does it identify drivers or read plates?
No. Stud Striker classifies tire tread — it isn’t built for facial recognition or license plate capture. The model only looks at the wheel/tread region of the frame.
Can we pull a report for a specific date, lane, or vehicle event?
Yes — every decision is timestamped and tagged with country, building, entry, and lane identifiers, so logs are searchable and auditable after the fact rather than relying on someone’s memory of a verbal check.
Who owns the data?
The property/operator. Detection logs, metrics, and archived frames live in your configured Loki, Prometheus, and Drive destinations — Stud Striker doesn’t retain a separate copy elsewhere.
Cost & Rollout
05How is pricing structured?
Per lane monitored — that maps directly to how the system is deployed: one camera, one config, one lane. A single lane is $399/month; multi-lane deployments are $299 per lane/month.
Can we pilot before committing property-wide?
Yes — the recommended path is a single-lane pilot to validate accuracy and fit for your site, then scale to additional lanes and buildings using the same configuration format.
What’s required from our team versus Kalkül’s?
Your team provides lane access, network connectivity, and camera mounting (or an existing feed). Kalkül handles configuration, model deployment, dashboard setup, and tuning the confidence threshold to your site’s conditions.
Support & Maintenance
06What happens if a camera or unit goes offline?
Live metrics (uptime, resource use, throughput) are visible on the Grafana dashboard per unit, so a stalled or offline camera is visible immediately rather than discovered days later during an audit.
Who do we contact for issues?
Kalkül provides setup, tuning, and ongoing support. Reach out at info@kalkul.ca or through your Kalkül engagement contact.
Can we add new metrics or integrations later?
Yes — the system is built as separate manager components (camera, model, config, metrics, upload, logging), so new integrations or metrics are additive and don’t require touching the rest of the system.
Still have a question?
Reach out and we'll help you scope a single-lane pilot for your property.
Email info@kalkul.ca