Recently, some state-of-the-art works have used deep learning-based architectures, specifically convolutional neural networks (CNNs), for banknote recognition and counterfeit detection with promising results. However, it is not clear which design strategy is more appropriate (custom or by transfer learning) in terms of classifier performance and inference times for massive data applications. This paper presents a comparison of the two design strategies in various types of architecture. For the transfer learning (TL) strategy, the most appropriate freezing points in CNN architectures (sequential, residual and Inception) are identified. In addition, a custom model based on an AlexNet-type sequential CNN is proposed. Both the TL and the custom models were trained and compared using a Colombian banknote dataset. According to the results, ResNet18 achieved the best accuracy, with 100%. On the other hand, the network with the shortest inference times was the proposed custom network, since its performance is up to 6.48-times faster in CPU and 16.29-times faster in GPU than the inference time with the models by transfer learning.

This article presents a system focused on the detection of three types of abnormal walk patterns caused by neurological diseases, specifically Parkinsonian gait, Hemiplegic gait, and Spastic Diplegic gait. A Kinect sensor is used to extract the Skeleton from a person during its walk, to then calculate four types of bases that generate different sequences from the 25 points of articulations that the Skeleton gives. For each type of calculated base, a recurrent neural network (RNN) is trained, specifically a Long short-term memory (LSTM). In addition, there is a graphical user interface that allows the acquisition, training, and testing of trained networks. Of the four trained networks, 98.1% accuracy is obtained with the database that was calculated with the distance of each point provided by the Skeleton to the Hip-Center point.

This paper presents the implementation of a DAG-CNN which aims to classify and verify the authenticity of the offline signatures of 3 users, using the writer-independent method. In order to develop this work, 2 databases (training / validation and testing) were built manually, i.e. the manual collection of the signatures of the 3 users as well as forged signatures made by people not belonging to the base and altered by the same users were done, and signatures of another 115 people were used to create the category of non-members. Once the network is trained, its validation and subsequent testing is performed, obtaining overall accuracies of 99.4% and 99.3%, respectively, showing the features learned by the network and verifying the ability of this configuration of neural network to be used in applications for identification and verification of offline signatures.