Fingerprint based Blood Group Detection

This paper provides a full-fleet of frame-based deep learning structure of blood group identification through fingerprint images. The data of the fingerprints of eight types of blood groups was in a dataset that included eight blood groups, namely, A, A-, B, B-, AB, AB-, O and O-, which was processed. The workflow contains data preparation, augmentation, transfer learning, model training, evaluation, as well as prediction. The four popular CNN architectures ResNet, VGG16, AlexNet, and LeNet were trained and evaluated in similar conditions. Accuracy, precision, recall and F1-score were used as the model performance measures.