The coordinates of a face, mapped and learned in Amsterdam.
We compute key points around eyes, nose, and mouth to train a neural network for accurate recognition. No photos saved, only anonymized vectors.
Face recognition is not magic. It is manual work.
You don't get perfect recognition from 'AI magic.' You get it from thousands of manual corrections, coordinate by coordinate, until the network stops guessing wrong.
Coordinate mapping
Eyes, nose, mouth — every facial coordinate computed from your photos and videos.
Two to four hours
Training runs slowly. We correct every error the network makes along the way.
Manual corrections
New networks make mistakes. We fix them by hand until accuracy holds.
Anonymized vectors only
Your photos and videos are processed and discarded. Only feature vectors remain.
Smart intercom ready
Trained networks integrate directly with smart intercom systems for fast, reliable entry.
Retrained on demand
If the network misreads a face, we keep the vector data and retrain the model.
Hundreds of training runs, zero shortcuts
Each training session takes two to four hours with manual corrections at every step. That level of precision delivers recognition that works when it counts.
We needed a system that could recognise residents even in low light and at awkward angles. After the two-hour training session with manual point adjustments, it worked. No false positives on the first night.
Lars van der Heijden
Property manager, Amsterdam residential complex
I was sceptical about the four-hour training window. But watching them manually correct each facial coordinate point was what convinced me. The accuracy after that single session has been consistent.
Fenna de Vries
Security systems integrator, Utrecht
No photos stored. That was non-negotiable for our tenants. The anonymised vector approach made it easy to get sign-off from our privacy board. And the facial mapping works exactly as described.
Ahmed Al-Rashid
Operations director, North Holland housing cooperative
Built in Amsterdam. Trained on coordinate precision.
We started with a small team in Amsterdam, writing software that maps faces to vectors. Every model goes through manual corrections because accuracy matters more than speed.
Anonymized face vectors and training time
Ready to see what your face data can do?
We'll walk you through the process, answer your questions, and set up a training session. No pressure, just a clear next step.