Abstract
Facial expression recognition (FER) is currently one of the most active research topics due to its wide range of applications in the human-computer interaction field. An important part of the recent success of automatic FER was achieved thanks to the emergence of deep learning approaches. However, training deep networks for FER is still a very challenging task, since most of the available FER data sets are relatively small. Although transfer learning can partially alleviate the issue, the performance of deep models is still below of its full potential as deep features may contain redundant information from the pre-trained domain. Instead, we propose a novel end-to-end neural network architecture along with a well-designed loss function based on the strong prior knowledge that facial expressions are the result of the motions of some facial muscles and components. The loss function is defined to regularize the entire learning process so that the proposed neural network is able to explicitly learn expression-specific features. Experimental results demonstrate the effectiveness of the proposed model in both lab-controlled and wild environments. In particular, the proposed neural network provides quite promising results, outperforming in most cases the current state-of-the-art methods.

Abstract

Abstract
Fingerprint segmentation is a crucial step of an automatic fingerprint identification system, since an accurate segmentation promote both the elimination of spurious minutiae close to the foreground boundaries and the reduction of the computation time of the following steps. In this paper, a new, and more robust fingerprint segmentation algorithm is proposed. The main novelty is the introduction of a more robust binarization process in the framework, mainly based on the fuzzy C-means clustering algorithm. Experimental results demonstrate significant benchmark progress on three existing FVC datasets.

Abstract

Abstract
The extraction of morphometric features from images of biological structures is a crucial task for the study of several diseases. Particularly, concerning neuropathies, the state of the myelination process is vital for neuronal integrity and may be an indicator of the disease type and state. Few approaches exist to automatically analyse nerve morphometry and assist researchers in this time consuming task. The aim of this work is to develop an algorithm to detect axons and myelin contours in myelinated fibres of sciatic nerve images, thus allowing the automated assessment and quantification of myelination through the measurement of the g-ratio. The application of a directional gradient together with an active contour algorithm was able to effectively and accurately determine the degree of myelination in an imagiological dataset of sciatic nerves. It was obtained an average error of 1.80%, in comparison with the manual annotation performed by the specialist in all dataset. © 2014 IEEE.