This is the official repository of the paper: Fusion-based Few-Shot Morphing Attack Detection and Fingerprinting, including codes, data and pre-trained model.
This is an implementation of few-shot learning based morphing attacks detection model, and is extended from binary detection to multiclass fingerprinting. The model aims at learning discriminative features which can be generalized to unseen morphing attack types from predefined presentation attacks. A large-scale face morphing database is collected, which contains 5 face subdatasets and 8 different morphing algorithms, to benchmark the proposed method.
paper link: contact:
5/31: create the project.
face detection and cropping: code (ref: https://github.com/Practical-CV/Facial-Landmarks-Detection-with-DLIB )
(1) PRNU: https://dde.binghamton.edu/download/camera_fingerprint/ (2) Noiseprint ('Noiseprint: a CNN-based camera model fingerprint') : https://github.com/grip-unina/noiseprint (3) FBC ('Revisiting Bilinear Pooling: A Coding Perspective'): https://ojs.aaai.org/index.php/AAAI/article/view/5811 (4) APL ('Adaptive Posterior Learning: few-shot learning with a surprise-based memory module'): https://github.com/cogentlabs/apl
If you use these codes, please cite their papers!