Improving Sensor Interoperability between Contactless and Contact-Based Fingerprints Using Pose Correction and Unwarping

Figure 2

The figure shows the architecture of a custom U-Net model used for both finger segmentation and core detection. The model consists of several convolutional layers with leaky ReLU activations (right-banded orange), a pretrained MobileNet V2 (light orange), and several transpose convolutional layers (blue). The intermediate outputs of the MobileNet V2 (right-banded blue) are used for skip connections, and concatenation is denoted by green spheres with a “+” symbol. Batch normalization layers are depicted as thin red layers following the transpose convolution blocks. The first violet layer is a resizing layer, and the second violet layer represents the Sigmoid/Softmax output activation.