Abstract
Cancer is one of the most serious health problems of our time. One approach for automatically classifying tumor samples is to analyze derived molecular information. Previous work by Teixeira et al. compared different methods of Data Oversampling and Feature Reduction, as well as Deep (Stacked) Denoising Autoencoders followed by a shallow layer for classification. In this work, we compare the performance of 6 different types of Autoencoder (AE), combined with two different approaches when training the classification model: (a) fixing the weights, after pretraining an AE, and (b) allowing fine-tuning of the entire network. We also apply two different strategies for embedding the AE into the classification network: (1) by only importing the encoding layers, and (2) by importing the complete AE. Our best result was the combination of unsupervised feature learning through a single-layer Denoising AE, followed by its complete import into the classification network, and subsequent fine-tuning through supervised training, achieving an F1 score of 99.61% +/- 0.54. We conclude that a reconstruction of the input space, combined with a deeper classification network outperforms previous work, without resorting to data augmentation techniques.

Abstract

Abstract
Developing robust and universal methods for unsupervised segmentation of moving objects in video sequences has proved to be a hard and challenging task that has attracted the attention of many researchers over the last decades. State-of-the-art methods are, in general, computationally heavy preventing their use in real-time applications. This research addresses this problem by proposing a robust and computationally efficient method, coined BMOG, that significantly boosts the performance of a widely used method based on a Mixture of Gaussians. The proposed solution explores a novel classification mechanism that combines color space discrimination capabilities with hysteresis and a dynamic learning rate for background model update. The complexity of BMOG is kept low, proving its suitability for real-time applications. BMOG was objectively evaluated using the ChangeDetection.net 2014 benchmark. An exhaustive set of experiments was conducted, and a detailed analysis of the results, using two complementary types of metrics, revealed that BMOG achieves an excellent compromise in performance versus complexity.

Abstract
We present a novel model for the characterization of musical rhythms that is based on the pervasive rhythmic phenomenon of syncopation. Syncopation is felt when the sensation of the regular beat or pulse in the music is momentarily disrupted; the feeling arises from breaking more expected patterns such as pickups (anacrusis) and faster events that introduce and bridge the notes articulated on the beats. Our model begins with a simple pattern that articulates a beat consistent with the metrical expectations of a listener. Any rhythm is then generated from a unique combination of transformations applied on that simple pattern. Each transformation introduces notes in off-beat positions as one of three basic characteristic elements: (1) syncopations, (2) pickup rhythmic figures and (3) faster notes that articulate a subdivision of the beat. The characterization of a pattern is based on an algorithm that discovers and reverses the transformations in a stepwise manner. We formalize the above transformations and present the characterization algorithm, and then demonstrate and discuss the model through the analysis of the main rhythmic pattern of the song Don't stop till you get enough' by Michael Jackson.

Abstract
Breast cancer is one of the most common malignanciesaffecting women worldwide. However, despite its incidence trends have increased, the mortality rate has significantly decreased. The primary concern in any cancer treatment is the oncological outcome but, in the case of breast cancer, the surgery aesthetic result has become an important quality indicator for breast cancer patients. In this sense, an adequate surgical planning and prediction tool would empower the patient regarding the treatment decision process, enabling a better communication between the surgeon and the patient and a better understanding of the impact of each surgical option. To develop such tool, it is necessary to create complete 3D model of the breast, integrating both inner and outer breast data. In this review, we thoroughly explore and review the major existing works that address, directly or not, the technical challenges involved in the development of a 3D software planning tool in the field of breast conserving surgery.

Abstract
Endoscopic capsules are vitamin-sized devices that create 8 to 10 hour videos of the digestive tract. They are the leading diagnosing method for the small bowel, a region not easily accessible with traditional endoscopy techniques. However, these capsules do not provide localization information, even though it is crucial for the diagnosis, follow-ups and surgical interventions. Currently, the capsule localization is either estimated based on scarce gastrointestinal tract landmarks or given by additional hardware that causes discomfort to the patient and represents a cost increase. Current software methods show great potential, but still need to improve in order to overcome their limitations. In this work, a visual odometry method for capsule localization inside the small bowel is proposed.