ABSTRACT
Even though it is among the top ten European wine producers and the top 20 worldwide, in Romania, the wine culture is not as well developed as in countries such as Portugal, Italy and France, considered the largest consuming countries of wine, from the European Union. In addition, there is a differentiation between urban and rural preferences for wine consumption. Romania is one of the major wine producers, at the European level, ranking 6th, with a total production of 5.4 million hl in 2021, being on an upward trend in last years. The national wine market is in a continuous rise, being determined by a growing demand, supported by the health benefits of wine products. The quality of the wines and the way they are presented, the innovation of flavors and the new distribution networks, combined with the changes that occurred during the Covid-19 pandemic, impose new consumption models.
ABSTRACT
Nowadays, reliable medical diagnostics from computed tomography (CT) and X-rays can be obtained by using a large number of image edge detection methods. One technique with a high potential to improve the edge detection of images is ant colony optimization (ACO). In order to increase both the quality and the stability of image edge detection, a vector called pheromone sensitivity level, PSL, was used within ACO. Each ant in the algorithm has one assigned element from PSL, representing the ant’s sensibility to the artificial pheromone. A matrix of artificial pheromone with the edge information of the image is built during the process. Demi-contractions in terms of the mathematical admissible perturbation are also used in order to obtain feasible results. In order to enhance the edge results, post-processing with the DeNoise convolutional neural network (DnCNN) was performed. When compared with Canny edge detection and similar techniques, the sensitive ACO model was found to obtain overall better results for the tested medical images;it outperformed the Canny edge detector by 37.76%.
ABSTRACT
It has recently been shown that the interpretation by partial differential equations (PDEs) of a class of convolutional neural networks (CNNs) supports definition of architectures such as parabolic and hyperbolic networks. These networks have provable properties regarding the stability against the perturbations of the input features. Aiming for robustness, we tackle the problem of detecting changes in chest X-ray images that may be suggestive of COVID-19 with parabolic and hyperbolic CNNs and with domain-specific transfer learning. To this end, we compile public data on patients diagnosed with COVID-19, pneumonia, and tuberculosis, along with normal chest X-ray images. The negative impact of the small number of COVID-19 images is reduced by applying transfer learning in several ways. For the parabolic and hyperbolic networks, we pretrain the networks on normal and pneumonia images and further use the obtained weights as the initializers for the networks to discriminate between COVID-19, pneumonia, tuberculosis, and normal aspects. For DenseNets, we apply transfer learning twice. First, the ImageNet pretrained weights are used to train on the CheXpert dataset, which includes 14 common radiological observations (e.g., lung opacity, cardiomegaly, fracture, support devices). Then, the weights are used to initialize the network which detects COVID-19 and the three other classes. The resulting networks are compared in terms of how well they adapt to the small number of COVID-19 images. According to our quantitative and qualitative analysis, the resulting networks are more reliable compared to those obtained by direct training on the targeted dataset.