UBNet: Deep learning-based approach for automatic X-ray image detection of pneumonia and COVID-19 patients

BACKGROUND: Analysis of chest X-ray images is one of the primary standards in diagnosing patients with COVID-19 and pneumonia, which is faster than using PCR Swab method. However, accuracy of using X-ray images needs to be improved. OBJECTIVE: To develop a new deep learning system of chest X-ray images and evaluate whether it can quickly and accurately detect pneumonia and COVID-19 patients. METHODS: The developed deep learning system (UBNet v3) uses three architectural hierarchies, namely first, to build an architecture containing 7 convolution layers and 3 ANN layers (UBNet v1) to classify between normal images and pneumonia images. Second, using 4 layers of convolution and 3 layers of ANN (UBNet v2) to classify between bacterial and viral pneumonia images. Third, using UBNet v1 to classify between pneumonia virus images and COVID-19 virus infected images. An open-source database with 9,250 chest X-ray images including 3,592 COVID-19 images were used in this study to train and test the developed deep learning models. RESULTS: CNN architecture with a hierarchical scheme developed in UBNet v3 using a simple architecture yielded following performance indices to detect chest X-ray images of COVID-19 patients namely, 99.6%accuracy, 99.7%precision, 99.7%sensitivity, 99.1%specificity, and F1 score of 99.74%. A desktop GUI-based monitoring and classification system supported by a simple CNN architecture can process each chest X-ray image to detect and classify COVID-19 image with an average time of 1.21 seconds. CONCLUSION: Using three hierarchical architectures in UBNet v3 improves system performance in classifying chest X-ray images of pneumonia and COVID-19 patients. A simple architecture also speeds up image processing time.

On the Hierarchical Use of Colourants in a 15th Century Book of Hours

An illuminated Book of Hours (in use in Chalon-sur-Saône) currently owned by the Museo Civico di Arte Antica and displayed in the prestigious Palazzo Madama in Torino (Italy) was investigated by means of optical microscopy, fibre optic reflectance spectroscopy, fibre optic molecular fluorimetry, X-ray fluorescence spectrometry and Raman spectroscopy. The aim of the scientific survey was to expand the knowledge of the manuscript itself and on the materials and techniques employed by Antoine the Lonhy, the versatile itinerant artist who decorated the book in the 15th century. The focus was to reveal the original colourants and to investigate the pigments used in rough retouches which were visible in some of the miniatures. The investigation was carried out in situ by portable instruments according to a non-invasive analytical sequence previously developed. It was evident that the use of different pigments by the master was ruled, at least partially, by a hierarchical scheme in which more precious materials were linked to the most important characters or details in the painted scene.

Hierarchical Scheme for Vehicle Make and Model Recognition

A vehicle make and model recognition (VMMR) system is a common requirement in the field of intelligent transportation systems (ITS). However, it is a challenging task because of the subtle differences between vehicle categories. In this paper, we propose a hierarchical scheme for VMMR. Specifically, the scheme consists of (1) a feature extraction framework called weighted mask hierarchical bilinear pooling (WMHBP) based on hierarchical bilinear pooling (HBP) which weakens the influence of invalid background regions by generating a weighted mask while extracting features from discriminative regions to form a more robust feature descriptor; (2) a hierarchical loss function that can learn the appearance differences between vehicle brands, and enhance vehicle recognition accuracy; (3) collection of vehicle images from the Internet and classification of images with hierarchical labels to augment data for solving the problem of insufficient data and low picture resolution and improving the model’s generalization ability and robustness. We evaluate the proposed framework for accuracy and real-time performance and the experiment results indicate a recognition accuracy of 95.1% and an FPS (frames per second) of 107 for the framework for the Stanford Cars public dataset, which demonstrates the superiority of the method and its availability for ITS.

Formation Tracking Control and Obstacle Avoidance of Unicycle-Type Robots Guaranteeing Continuous Velocities

In this paper, we addressed the problem of controlling the position of a group of unicycle-type robots to follow in formation a time-varying reference avoiding obstacles when needed. We propose a kinematic control scheme that, unlike existing methods, is able to simultaneously solve the both tasks involved in the problem, effectively combining control laws devoted to achieve formation tracking and obstacle avoidance. The main contributions of the paper are twofold: first, the advantages of the proposed approach are not all integrated in existing schemes, ours is fully distributed since the formulation is based on consensus including the leader as part of the formation, scalable for a large number of robots, generic to define a desired formation, and it does not require a global coordinate system or a map of the environment. Second, to the authors’ knowledge, it is the first time that a distributed formation tracking control is combined with obstacle avoidance to solve both tasks simultaneously using a hierarchical scheme, thus guaranteeing continuous robots velocities in spite of activation/deactivation of the obstacle avoidance task, and stability is proven even in the transition of tasks. The effectiveness of the approach is shown through simulations and experiments with real robots.

Physical-geographical natural systems within waters of the World Ocean

Based on the data of theoretical developments in the fields of ocean geography and system-geographical analysis, a hierarchical scheme of natural systems in the water layer of the World Ocean has been examined. The aim of the work is to carry out the first attempt to compare landscapes on land, natural systems in the coastal zone (the zone of contact between land and the World Ocean) and those in the World Ocean. The differentiation of the oceanic natural environment which is a possible version of a systematised list of systems ranging from individual oceans to individual eddies (or impulses) in the deep sea and on the shelf of shallow water are discussed. This work therefore, attempts to find new ways for the synchronous study of the hierarchical series of the coastal zone and the water layer of the World Ocean, along with land landscapes as part of the geographic shell of the Earth. This approach will make it possible to obtain a series of systems for the entire geographic envelope. This is a promising approach for an indebt development of physical geography in general.

Crystal structure relations in the binary system Li–Sn including the compound c-Li3Sb

A topological analysis of the crystal structures of Li, Li–Sn compounds, Li8Sn3−xSbx and metastable c-Li3Sb showed that these structures can be described by a hierarchical scheme of building blocks based on atom blocks and polyhedra blocks, respectively. These blocks are linked in distinct ways to form the individual 3D atom arrangement. A common model was established for the construction of the mentioned structures from bespoke building blocks, for which bcc-Li is the aristotype. This latter structure can be described on the basis of hexa-capped cubes from which variants are derived through substitution of Li by Sn (or Sb). These are then combined into polyhedra blocks that are in turn assembled into polyhedra sequences. These latter are repeated and linked in three dimensions to form the whole crystal structure. At xSn ≥ 0.5, this mechanism changes and structural elements from bcc-Li and β-Sn can be observed in LiSn and Li2Sn5. In this work, we present the similarities and differences between the various crystal structures, the topological model with its construction rules and its limitations.

Unsupervised 3D Motion Summarization Using Stacked Auto-Encoders

In this paper, a deep stacked auto-encoder (SAE) scheme followed by a hierarchical Sparse Modeling for Representative Selection (SMRS) algorithm is proposed to summarize dance video sequences, recorded using the VICON Motion capturing system. SAE’s main task is to reduce the redundant information embedding in the raw data and, thus, to improve summarization performance. This becomes apparent when two dancers are performing simultaneously and severe errors are encountered in the humans’ point joints, due to dancers’ occlusions in the 3D space. Four summarization algorithms are applied to extract the key frames; density based, Kennard Stone, conventional SMRS and its hierarchical scheme called H-SMRS. Experimental results have been carried out on real-life dance sequences of Greek traditional dances while the results have been compared against ground truth data selected by dance experts. The results indicate that H-SMRS being applied after the SAE information reduction module extracts key frames which are deviated in time less than 0.3 s to the ones selected by the experts and with a standard deviation of 0.18 s. Thus, the proposed scheme can effectively represent the content of the dance sequence.