We propose a structured prediction architecture, which exploits the local generic features extracted by Convolutional Neural Networks and the capacity of Recurrent Neural Networks (RNN) to retrieve distant dependencies. The proposed architecture, called ReSeg, is based on the recently introduced ReNet model for image classification. We modify and extend it to perform the more challenging task of semantic segmentation. Each ReNet layer is composed of four RNN that sweep the image horizontally and vertically in both directions, encoding patches or activations, and providing relevant global information. Moreover, ReNet layers are stacked on top of pre-trained convolutional layers, benefiting from generic local features. Upsampling layers follow ReNet layers to recover the original image resolution in the final predictions. The proposed ReSeg architecture is efficient, flexible and suitable for a variety of semantic segmentation tasks. We evaluate ReSeg on several widely-used semantic segmentation datasets: Weizmann Horse, Oxford Flower, and CamVid; achieving stateof-the-art performance. Results show that ReSeg can act as a suitable architecture for semantic segmentation tasks, and may have further applications in other structured prediction problems. The source code and model hyperparameters are available on https://github.com/fvisin/reseg.

Lately, with deep learning outpacing the other machine learning techniques in classifying images, we have witnessed a growing interest of the remote sensing community in employing these techniques for the land use and land cover classification based on multispectral and hyperspectral images; the number of related publications almost doubling each year since 2015 is an attest to that. The advances in remote sensing technologies, hence the fast-growing volume of timely data available at the global scale, offer new opportunities for a variety of applications. Deep learning being significantly successful in dealing with Big Data, seems to be a great candidate for exploiting the potentials of such complex massive data. However, there are some challenges related to the ground-truth, resolution, and the nature of data that strongly impact the performance of classification. In this paper, we review the use of deep learning in land use and land cover classification based on multispectral and hyperspectral images and we introduce the available data sources and datasets used by literature studies; we provide the readers with a framework to interpret the-state-of-the-art of deep learning in this context and offer a platform to approach methodologies, data, and challenges of the field.

No abstract

This paper presents a method for obstructive sleep apnea (OSA) screening based on the electrocardiogram (ECG) recording during sleep. OSA is a common sleep disorder produced by repetitive occlusions in the upper airways and this phenomenon can usually be observed also in other peripheral systems such as the cardiovascular system. Then the extraction of ECG characteristics, such as the RR intervals and the area of the QRS complex, is useful to evaluate the sleep apnea in noninvasive way. In the presented analysis, 50 recordings coming from the apnea Physionet database were used; data were split into two sets, the training and the testing set, each of which was composed of 25 recordings. A bivariate time-varying autoregressive model (TVAM) was used to evaluate beat-by-beat power spectral densities for both the RR intervals and the QRS complex areas. Temporal and spectral features were changed on a minute-by-minute basis since apnea annotations where given with this resolution. The training set consisted of 4950 apneic and 7127 nonapneic minutes while the testing set had 4428 apneic and 7927 nonapneic minutes. The K-nearest neighbor (KNN) and neural networks (NN) supervised learning classifiers were employed to classify apnea and non apnea minutes. A sequential forward selection was used to select the best feature subset in a wrapper setting. With ten features the KNN algorithm reached an accuracy of 88%, sensitivity equal to 85%, and specificity up to 90%, while NN reached accuracy equal to 88%, sensitivity equal to 89% and specificity equal to 86%. In addition to the minute-by-minute classification, the results showed that the two classifiers are able to separate entirely (100%) the normal recordings from the apneic recordings. Finally, an additional database with eight recordings annotated as normal or apneic was used to test again the classifiers. Also in this new dataset, the results showed a complete separation between apneic and normal recordings.

We describe a system for the evaluation of the sleep macrostructure on the basis of Emfit sensor foils placed into bed mattress and of advanced signal processing. The signals on which the analysis is based are heart-beat interval (HBI) and movement activity obtained from the bed sensor, the relevant features and parameters obtained through a time-variant autoregressive model (TVAM) used as feature extractor, and the classification obtained through a hidden Markov model (HMM). Parameters coming from the joint probability of the HBI features were used as input to a HMM, while movement features are used for wake period detection. A total of 18 recordings from healthy subjects, including also reference polysomnography, were used for the validation of the system. When compared to wake-nonrapid-eye-movement (NREM)-REM classification provided by experts, the described system achieved a total accuracy of 79+/-9% and a kappa index of 0.43+/-0.17 with only two HBI features and one movement parameter, and a total accuracy of 79+/-10% and a kappa index of 0.44+/-0.19 with three HBI features and one movement parameter. These results suggest that the combination of HBI and movement features could be a suitable alternative for sleep staging with the advantage of low cost and simplicity.