This pretrained model is designed for finger vein recognition. It uses a MobileNet-based feature extractor trained on finger images to extract deep biometric features.
🔧 How It Works:
The model first extracts features from finger vein images using MobileNet.
These features are then used to form image pairs.
A deep neural network (e.g. Siamese) is trained on these pairs to learn a similarity metric.
Finally, the system classifies whether two finger vein images belong to the same person or not.
📦 Use Cases:
🔐 Biometric authentication systems
🔍 Finger vein matching or verification
🧬 Medical/Forensic identification tasks
🖼️ Input:
RGB finger vein image (resized to 224×224)
Normalized to [0, 1]
📤 Output:
Feature vector (if using encoder only)
Or: Match / No-match decision (in Siamese setup)
💾 Model Format:
model.keras — Keras format for MobileNet feature extractor
💾 code Licence:
Alaerjan, A.S., Mostafa, A.M., Mahmoud, A.A. et al. Efficient multi-finger vein recognition using layer-wise progressive MobileNet fine-tuning and a Dense-Head Probabilistic Siamese Network. Sci Rep (2025).
https://doi.org/10.1038/s41598-025-32132-5
