Identity from the veins on the back of a hand, fully offline

What it does

Verifies who someone is from the vein pattern on the back of their hand, entirely on-device. No network, no server holding biometrics.

How

A MobileNetV2 feature extractor with Siamese and triplet matchers, exported to TFLite so it runs on the device itself.

Evaluation is subject-disjoint: no person appears in both training and test data, so the scores reflect performance on genuinely unseen people rather than memorised individuals. That distinction is the difference between a number that means something and one that does not.

Data

I built a custom Arduino-controlled infrared capture rig to collect the dataset, since nothing suitable existed.