Search

Security is an essential condition for the well-being of human beings, which is why society seeks measures capable of preventing and identifying people who commit attacks against its security. One of the security solutions relied on video surveillance systems since through them scenes can be viewed directly and comfortably, through the Internet, using a computer, or any other mobile device. Nevertheless, this solution is not perfect because one of the most important limitations is related to the quality of the images acquired by these systems. They can be low resolution and can have changes in light intensity such as the clarity or darkness of the image, image perspective, and scale, which makes it difficult to recognize the people present in the room, even by human beings. Nowadays with the usage of Deep Learning algorithms, one can improve the quality of the images, which means, passing low to super-resolution images with relatively low effort. Nevertheless, the impact of this improvement is not analyzed to maintain the identity of the faces or person recognition task. This work proposes an artificial vision system for the improvement of faces in low-resolution images acquired through low-quality video surveillance systems. The work considers all the stages of a traditional artificial vision system, allowing the entry of images from a computer file, and improving the image through the super-resolution Enhance Deep Residual Networks algorithm, subsequently locates the person’s face using Mediapipe algorithm and performs the reconstruction only of the face area. Later, the impact of how good the performance of a face recognition model is taking advantage of the improvement image methods. We analyzed how much the enhancement procedure improves the overall face recognition system. We conduct our experiments with four image improvement algorithms: EDSR, ESPCN, FSRCNN, and LapSRN. Also, we select two standard datasets: 3DPeS and ChokePoint. Also, we did a final face recognition phase, where the LFW (Labeled Faces in the Wild) dataset was used to train and test a face recognition model. Related to image improvement, the results show that our proposal significantly visually improves the image more than quantitatively. Regarding the preservation of the face identity, the face recognition model achieves an accuracy of 94% by using images with enhancements and reconstructions in the facial area which is slightly lower than using the original images. Although considerable improvements were achieved with several super-resolution and enhancement models, it is crucial to analyze in which measure this improvement damages the face recognition model. Spite we achieved a very similar recognition rate comparing the improvement and not-improvement images, more analysis must be explored to go deeper into this conclusion. [ABSTRACT FROM AUTHOR]

Science as a whole is organized into broad fields, and as a consequence, research, resources, students, etc., are also classified, assigned, or invited following a similar structure. Some fields have been established for centuries, and some others are just flourishing. Funding, staff, etc., to support fields are offered if there is some activity on it, commonly measured in terms of the number of published scientific papers. How to find them? There exist well-respected listings where scientific journals are ascribed to one or more knowledge fields. Such lists are human-made, but the complexity begins when a field covers more than one area of knowledge. How to discern if a particular paper is devoted to a field not considered in such lists? In this work, we propose a methodology able to classify the universe of papers into two classes; those belonging to the field of interest, and those that do not. This proposed procedure learns from the title and abstract of papers published in monothematic or "pure" journals. Provided that such journals exist, the procedure could be applied to any field of knowledge. We tested the process with Geographic Information Science. The field has contacts with Computer Science, Mathematics, Cartography, and others, a fact which makes the task very difficult. We also tested our procedure and analyzed its results with three different criteria, illustrating its power and capabilities. Interesting findings were found, where our proposed solution reached similar results as human taggers also similar results compared with state-of-the-art related work. [ABSTRACT FROM AUTHOR]

Spanish is one of the most spoken languages in the world. Its proliferation comes with variations in written and spoken communication among different regions. Understanding language variations can help improve model performances on regional tasks, such as those involving figurative language and local context information. This manuscript presents and describes a set of regionalized resources for the Spanish language built on 4-year Twitter public messages geotagged in 26 Spanish-speaking countries. We introduce word embeddings based on FastText, language models based on BERT, and per-region sample corpora. We also provide a broad comparison among regions covering lexical and semantical similarities and examples of using regional resources on message classification tasks. [ABSTRACT FROM AUTHOR]

Intelligent systems have been developed for years to solve specific tasks automatically. An important issue emerges when the information used by these systems exhibits a dynamic nature and evolves. This fact adds a level of complexity that makes these systems prone to a noticeable worsening of their performance. Thus, their capabilities have to be upgraded to address these new requirements. Furthermore, this problem is even more challenging when the information comes from human individuals and their interactions through language. This issue happens more easily and forcefully in the specific domain of Sentiment Analysis, where feelings and opinions of humans are in constant evolution. In this context, systems are trained with an enormous corpus of textual content, or they include an extensive set of words and their related sentiment values. These solutions are usually static and generic, making their manual upgrading almost unworkable. In this paper, an automatic and interactive coaching architecture is proposed. It includes a ML framework and a dictionary-based system both trained for a specific domain. These systems converse about the outcomes obtained during their respective learning stages by simulating human interactive coaching sessions. This leads to an Active Learning process where the dictionary-based system acquires new information and improves its performance. More than 800, 000 tweets have been gathered and processed for experiments. Outstanding results were obtained when the proposed architecture was used. Also, the lexicon was updated with the prior and new words related to the corpus used which is important to reach a better sentiment analysis classification. [ABSTRACT FROM AUTHOR]

A novel method for person identification based on soft-biometrics and oriented to work in real video surveillance environments is proposed in this paper. Thus, an evaluation of relevance's level of several appearance features is carried out with this purpose. First, a bag-of-soft-biometric features related to color, texture, local features, and geometry are extracted from individuals. The relevance of each feature has been deeply analyzed through different proposed methods. Features are ranked and weighted according to their relevance value. Later, each method is evaluated under two different scenarios: mono-camera and multi-camera surveillance images. In order to test the system in a realistic way, it has been evaluated over standard databases in the surveillance community: PETS 2006, PETS 2009, CAVIAR, SAIVT-SoftBio, and CAVIAR4REID. Moreover, a new database was acquired at Adolfo Suarez Madrid-Barajas international airport. This database was acquired under regular conditions and infrastructure of the Barajas airport, no additional camera or special settings were installed for this purpose. An analysis of relevance for each feature acquired in these two scenarios is presented. The results obtained demonstrate the promising potential of the soft-biometric approach. Finally, an optimal system configuration according to each scenario is obtained. [ABSTRACT FROM AUTHOR]