APPLICATION OF XCEPTION ARCHITECTURE FOR DEEP LEARNING-BASED FINGERPRINT RECOGNITION

This study proposes a convolutional neural network (CNN) method with an xception architecture model that is used to classify the types of fingerprint image patterns. The data used in this study uses data taken directly using a scanning tool made using FPM 10 A sensors and Arduino Uno. The dataset consists of five types of fingerprint image patterns, namely arch, ulnar loop, whorl, radial loop and twinted loop with a total of 1000 data. The research started from data collection, pre-processing, CNN architecture design, model training and evaluation. The application of the xception architecture shows the best performance with high test accuracy values, stable and consistent. The test scenario of this study is to compare different epoch values, namely 10.30 and 50 and use two learning rates, namely 0.0001 and 0.001. The best test results were obtained at epoch 30 with a learning rate of 0.0001, which was 92% and 93% at epoch 50 with a learning rate of 0.001.