Advanced face recognition. Research-grade accuracy. No GPU. Deployable anywhere.
Show and Go is SirusAI's flagship face recognition system — a state-of-the-art model that delivers results most commercial systems cannot match, at a fraction of the hardware cost. It runs on a standard CPU, handles the environments that break other systems, and applies to any use case that requires reliable, fast, on-device face recognition.
Show and Go is not an attendance product. It is a face recognition system — capable of identifying people accurately in conditions that defeat most commercial alternatives: low light, high camera angles, low resolution, moving subjects. And it does all of this on a standard CPU with no cloud dependency.
The intelligence behind it is srs-face-v2, SirusAI's proprietary model built from original research and field-validated across dozens of real-world deployments. It is accurate enough to trust, fast enough to operate at scale, and lean enough to run on hardware you already own.
What you build on top of it is entirely up to your use case. Attendance is one application. Access control, security monitoring, visitor management, patient identification, event check-in — the engine handles all of it.
98.4% top-1 identification on 480p CCTV streams — +17 points above the published baseline for the same conditions. Measured, documented, reproducible.
The most expensive barrier to deploying face recognition at scale is hardware. srs-face-v2 is quantized and optimised to run on a single CPU core at 12 ms per inference. One box handles 16 simultaneous streams.
Low light, steep angles, 480p resolution, moving subjects — we trained on those conditions, not curated lab datasets. The model performs where others stop working.
No cloud round-trip. No data leaving the premises. Detection, recognition, and anti-spoof all run locally. Results in milliseconds, privacy intact.
Six capabilities that define srs-face-v2. Each is backed by benchmarks and field data. Ask for the methodology and we will share it.
Identifies subjects accurately two stops below the threshold where the previous baseline breaks down. Trained on factory night shifts, dimly lit corridors, and unlit car parks.
No GPU, no accelerator card, no cloud. Fully quantized to int8 with no measurable accuracy loss. Runs on the same class of hardware already in most server rooms.
Ceiling-mounted cameras, high-angle CCTV, partial profiles — trained on the angles you actually encounter in the field, not the flattering portrait angles that look good on a demo slide.
A second model rejects printed photographs, video replays on a phone, and 3D masks — all within 4 ms after every match. No cloud round-trip.
Biometric data, face embeddings, and match logs stay on the device inside your building. There is no SirusAI cloud in the live path. Your data is yours, fully.
Structured match events, signed webhooks, REST output — the system pushes results to whatever your application needs. HRMS, ERP, access control, custom dashboards.
Show and Go is a face recognition engine. The application is yours to define. These are the categories where srs-face-v2 is already deployed or ready to deploy — each with distinct requirements the model meets.
Mark presence automatically as people pass a camera — no badge, no fingerprint, no queue. The roster updates itself. Works for daily attendance in schools and factories, event check-in, and lecture roll-calls without disrupting the flow of movement.
Authenticate identity at entry points and trigger access decisions in real time. Flag unknown faces, restrict zones by identity, and build a full audit trail — all without replacing existing CCTV infrastructure.
Identify patients at intake without physical cards or tokens. Enforce staff-only zones. Build an audit trail for regulatory compliance. The on-device model ensures sensitive biometric data never leaves your premises.
Run continuous recognition across a camera network — identifying known persons, logging movement, or tracking presence across locations. Designed to operate reliably in low-visibility conditions where most systems degrade.
If your application needs face recognition but building a model from scratch isn't your core work, srs-face-v2 is available as an API. Integrate accurate, field-proved recognition into any product — without owning the research infrastructure behind it.
If your requirement doesn't map to a category above, bring it to us. We build custom AI products around specific problems. If face recognition is part of your solution, we scope it, probe it, and build it — with the same research discipline that produced srs-face-v2.
srs-face-v2 against the published state-of-the-art baseline across the conditions that matter most in real deployments. Datasets and full methodology available on request.
Show and Go sits between your camera input and your application layer. It ingests streams, runs recognition, and pushes structured results downstream. The inference box is the only new hardware — everything else is yours.
We assess existing cameras and pull footage for calibration. We tell you before anything is signed whether the system fits your environment.
One week in our lab. We tune srs-face-v2 to your specific camera angles, lighting conditions, and subject demographics. You receive a calibration report with measured accuracy.
One afternoon. We bring the inference box, connect it to your camera streams, load the subject database, and verify all streams are live. Running by end of day.
Weekly reviews for the first month. We examine false matches and false rejects together and tighten the model if needed. Then quarterly — indefinitely.
Anonymised field reports from live deployments. Names withheld at the client's request. Raw numbers and full methodology available under NDA.
Replaced an RFID badge system with Show and Go on two existing CCTV cameras at the gate. Students walk in; the roster updates. First-period absentees identified by 8:35 a.m., parents notified by 8:40.
Three existing CCTV cameras at the canteen entrance. Shift attendance and contractor sign-in unified into one ledger. Night-shift accuracy held above 94% for six consecutive months.
No reader inside the classroom. Students are recognised as they walk the corridor; the system maps each face to the lecture they belong to. Lecturers receive a real-time roster on their phone.
No. Show and Go is a face recognition system. Attendance is the most common deployment we have today, but the same engine handles access control, visitor management, patient identification, security monitoring, and any other use case that needs reliable face recognition. Tell us your application and we will tell you honestly whether it fits.
One Intel NUC — no GPU, no accelerator, no cloud. The same class of box that already runs DVR software in most facilities. One box handles up to 16 simultaneous CCTV streams. If you need more, we add boxes. It scales linearly.
Most commercial APIs were designed for cooperative subjects in good light. srs-face-v2 was designed and field-validated for CCTV conditions: distance, angle, poor light, 480p resolution. Our benchmarks show a 17-point accuracy advantage over the published baseline on those conditions. We are also fully on-device — no cloud cost, no data leaving the premises.
Nowhere. All face embeddings, match logs, and biometric data stay on the inference box inside your building. There is no SirusAI cloud in the live inference path. We will sign whatever data processing agreement your legal team requires.
Yes. srs-face-v2 is available as an API for developers and businesses who want to embed face recognition into their own product without building the model themselves. See the API page for documentation, or contact us to discuss licensing.
Two weeks from a signed agreement. One day on-site for camera assessment, one week to calibrate to your environment, one afternoon to install and go live. We hand off a written runbook so your team can handle day-to-day operations — and site number two — without us.
Research-grade accuracy at a fraction of what comparable systems require in hardware and cloud spend. Tell us your use case and we will tell you honestly whether Show and Go fits.
