TrustWatch is an AI-powered web application that authenticates luxury watches from a single photo of the dial, delivering a verdict with calibrated confidence scores, attention heatmaps, and an AI-written narrative report.
What is TrustWatch?
TrustWatch is a decision-support system for luxury watch authentication, developed by Hoken Tech. It takes a clear photograph of the watch dial (minimum 500×500 px) and analyzes it using an EfficientNetV2 classifier. The output includes a predicted class (Real, Fake, or Not-a-watch), a recalibrated confidence score, a Diffuse Grad-CAM heatmap highlighting the regions that influenced the decision, and a written summary. The platform runs entirely in the browser, with images resized client-side and not retained beyond the request.
Key Features
- Easy photo upload – Users submit a clear dial photo; images are anonymized and resized client-side, never stored long-term.
- Rapid AI analysis – The onboard EfficientNetV2 classifier returns a verdict in under 3 seconds.
- Explainable results – A Diffuse Grad-CAM heatmap overlays the image, showing which regions (logos, indices, rehaut) drove the prediction.
- Calibrated confidence – Scores are re-normalized to filter out "Not-a-watch" noise, keeping the authenticity signal stable.
- Narrative report – An AI-generated summary translates the numeric scores into a client-ready paragraph; PDF export available on the Plus tier.
- Decision band – Ambiguous cases are automatically flagged for manual review, preventing overconfident false positives.
- Broad brand coverage – Supports over 500 models from eight luxury brands: Rolex, Richard Mille, IWC, Omega, Patek Philippe, Panerai, Audemars Piguet, and TAG Heuer.
- API for enterprise – Direct REST API with
/predictand/analyzeendpoints, rotating keys, volume-tiered quotas, and custom integration support.
Who is it for?
- Watch collectors and enthusiasts – Quickly verify a watch before purchasing, especially when buying second-hand online.
- Dealers and consultants – Authenticate multiple watches per month and generate PDF reports for clients.
- Jewellers and pre‑owned boutiques – Integrate TrustWatch via the API into their own systems for high‑volume screening.
What can you do with TrustWatch?
- Authenticate a watch before purchase – Upload a seller’s photo and get a verdict in seconds, avoiding costly fakes.
- Generate a shareable report – Download a PDF (Plus tier) with heatmap, confidence scores, and narrative to share with buyers or sellers.
- Flag suspicious listings – Use the decision band to identify watches that need expert inspection.
How does TrustWatch work?
The process consists of three steps: Upload, Analyse, and Explain. First, the user submits a clear photograph of the dial. The image is anonymized and resized client-side. Next, the AI model (EfficientNetV2) scores the image for Real, Fake, and Not-a-watch classes, with recalibrated confidence. Finally, the platform generates a Diffuse Grad-CAM attention heatmap and an AI-written narrative explaining which regions of the watch drove the verdict.
Pricing
TrustWatch offers two monthly subscription tiers and an enterprise plan. Personal – TrustWatch Ultimate costs €12/month and includes 30 watch scans and uploads per month. Connoisseur – TrustWatch Plus costs €15/month and provides 120 authentications per month, plus attention heatmaps, confidence scoring, AI-written narratives, PDF reports, and multi-method explainability. A free trial (no card required) is available for the first analysis. Enterprise API access is priced on request with volume-tiered quotas and SLAs.
FAQ
Does TrustWatch guarantee a watch is genuine?
No. TrustWatch is a decision-support tool, not a legal guarantee. It is designed for screening and review before a deeper manual inspection by an expert. High-value decisions should always combine AI results with expert examination.
What photo quality is needed?
A clear, well-lit image at least 500×500 pixels with the watch centred. Avoid heavy blur, extreme reflections, or cluttered backgrounds.
Do you store my photographs?
The platform is designed for in-memory processing. Images are resized client-side and are not retained beyond the request without explicit consent.
How do I cancel my subscription?
Cancellation is handled through the customer billing portal. Any changes take effect at the start of the next billing cycle.
