Anselmo Rafael Cukla

Accurate forecasting of digital storage usage is essential for capacity planning in judicial information systems, where cumulative growth and long-term temporal dependencies create significant operational challenges. This study evaluates neural network–based models for medium-term storage usage forecasting using real workload data from a production-grade object storage platform deployed in a judicial environment. The task is formulated as a univariate time series forecasting problem, adopting a differencing and cumulative reconstruction strategy to enable predictions on the absolute usage scale.Three neural network architectures are analyzed: a Multilayer Perceptron, a Long Short-Term Memory network, and a Gated Recurrent Unit, representing both feedforward and recurrent modeling approaches. All models are assessed under a unified preprocessing, training, and evaluation framework, considering point-wise accuracy, distributional consistency, preservation of temporal dependence, and long-horizon stability.Results show that convergence in the normalized training domain does not necessarily translate into reliable performance after reconstruction to the absolute scale. The feedforward architecture achieves the highest predictive accuracy and strongest statistical consistency, while recurrent models exhibit greater sensitivity to configuration choices and reduced fidelity in high-usage regimes. These findings emphasize the importance of preprocessing and feature representation and demonstrate that, under appropriate transformations, feedforward neural networks can outperform recurrent models for storage capacity forecasting in large-scale judicial systems.

Traditional Visual SLAM systems, like ORB-SLAM3, often lose accuracy in dynamic environments. This work presents YOLO11-ORB-SLAM3, an enhancement to ORB-SLAM3 for dynamic scenarios, which integrates a YOLO11-based instance segmentation module to detect and exclude dynamic features from the tracking process. The system is designed to work with stereo and RGB-D cameras, and its performance was evaluated on challenging dynamic sequences of the public TUM RGB-D dataset, and also through real-world experiments on a mobile robot using a stereo camera to highlight its robustness and viability for real robotic applications. Experimental results demonstrate that the proposed system outperforms the original ORB-SLAM3, reducing the error by 93% in the public TUM dataset while preserving computational efficiency.

This study presents the use of convolutional neural networks aimed at detecting and simultaneously translating continuous signs of Brazilian Sign Language (LIBRAS) in videos captured by common computer cameras. The objective of this work was to propose a methodology that enabled the classification of signs performed by people of different genders, body types, and skin tones, with minimal interference from the video background while using cameras of medium or low quality, creating an accessible approach for various applications. For this purpose, the MediaPipe algorithm was used for video data extraction, the FastDTW algorithm was employed for data standardization, convolutional neural networks were implemented using the TensorFlow and Keras libraries for sign recognition, and a custom dataset was created to train the network. All these tools were integrated using the Python programming language to produce a model for real-time classification of continuous LIBRAS signals.

This paper employs deep learning algorithms to address the classification of optical fiber cables through image processing techniques. The VGG-16 and Res-Net50 neural network architectures are utilized, and their performance is evaluated in comparison. The experiments were developed in an optical fiber dataset composed of optical fiber images and increased with synthetic data using the data augmentation technique; the dataset has optical fiber images in different states of preservation and colors. The neural networks will classify the images into two classes ("good" or "bad").

This article presents a methodology for estimating the gripping forces in a Fin Ray-type gripper, based on the integration of experimental and computational approaches. The development and validation methods includes (1) mechanical modeling and material selection; (2) experimental tests to relate FG finger displacement to maximum applied force using a load cell; (3) validation of the computational model through finite element method (FEM) simulations in ABAQUS using experimental data; and (4) experimental analysis of the FG handling a chicken egg, with the FEM determining the stress applied to the egg. The computational results showed a maximum stress of approximately 7 MPa on the egg, with no signs of damage, demonstrating the FG’s suitability for handling delicate objects in both the experimental and computational procedures, thus enabling safe object handling without causing damage. This work advances research on Fin Ray-type flexible end-effectors, emphasizing their utility in manipulating fragile objects without requiring complex force and pressure control algorithms.

Considerando os desafios enfrentados pela perícia criminal no Brasil, como a alta demanda por laudos, escassez de recursos e falta de padronização nos laudos de crimes contra o patrimônio, este trabalho objetivou-se a elencar gargalos de eficiência e analisar o processo de elaboração de Laudos Periciais na SPTC/GO sob a ótica do Lean Office para identificar desperdícios, propor melhorias no fluxo de trabalho e demonstrar como essas mudanças podem reduzir o backlog de exames e aumentar a eficiência institucional. Para tanto, realizou-se o mapeamento do processo para analisar o cenário atual, identificar gargalos para uma possível eliminação de atividades sem valor, reduzir o tempo de resposta e melhorar a alocação de recursos. Concluiu-se assim, que a aplicação dos princípios de Lean Thinking na Polícia Científica de Goiás se revela uma estratégia eficaz para reduzir desperdícios, otimizar e agregar valor ao processo de elaboração de Laudos Periciais.

In the last five years, around 150 bank robbers have died in clashes with Brazilian police. This wave of violence became known as “Novo Cangaço”. Despite the gangs using explosives, rifles and .50 machine guns, this wave of terror was not classified as terrorism in Brazil. The objective of this research is to investigate this type of organised crime and its similarity to the concept of global terrorism. This is a comprehensive review of how the “Novo Cangaço” (new bandit) is understood. We investigated around 18 thousand articles and highlighted four distinct groupings to explain what terrorism is in the modern world. The findings from interviews allowed us to understand that it is only possible to classify a terrorist act under current laws if there is an attack on public events or public facilities. We interviewed police and military personnel with experience in combatting this variant of modern horror. Our conclusions indicate that this wave of violence associated with drug trafficking should be classified as terrorism. Considering these points, we believe that the “Novo Cangaço” is a form of terrorism, given its connection with drug trafficking and political movements. The use of explosives, the siege of cities, the use of hostages and violence aimed at causing terror and commotion also reinforce this definition. Such theft practices may cease to exist due to the disappearance of paper money, but the legacy will remain and serve as a lesson for those who wish to commit similar crimes. We also emphasise that this type of crime could soon inspire other gangs to commit violent crimes in Latin America.

Between December 2019 and December 2022, COVID-19 caused approximately 670,000 deaths and overwhelmed health systems globally, amid significant misinformation. In this sense, interface analysis can assist in treatment and prevention, by assessing whether the information is reliable, clear, and objective. This work aims to analyze the interface of the Coronavírus SUS mobile application, focusing on epidemiological infectious diseases, which aims to raise awareness about communicable diseases, their prevention, and treatment. Metrics related to the development of the interface and human-computer interaction will be used, since ease of viewing, learning, and navigation of the application by the user is crucial for user engagement. The application, although simple, has many areas for improvement. As a result, it can be observed that there are inconsistencies in the “visibility of system status”, “user control” and “consistency” between screens, with “aesthetics and minimalist design” problems in the initial interface and “flexibility and usability” problems in the tips and news interfaces.

The integration of robotics into education is becoming increasingly widespread, offering effective tools for teaching. Extensive literature highlights the potential of robotics in programming education and in enhancing problem-solving skills through competitions. Educational games are also effective in improving memorization and analytical skills in mathematics and physics. This study utilizes a sniper robot to teach physics, focusing on projectile motion with and without air resistance. The robot’s mechanical, electrical, and software components enable practical learning. The methodology includes theoretical classes, tests, and hands-on experiments. The results present a teaching plan and suggested exercises, emphasizing the robot’s effectiveness in classroom settings. The conclusion underscores the educational benefits of utilizing the launch robot in physics instruction.

Recentemente, sistemas de recomendação se expandiram para a área da educação. Sistemas que sugerem itens de interesse, ao explorar as preferências dos alunos, ajudam a lidar com a sobrecarga de informações. Este trabalho tem como objetivo desenvolver um sistema de recomendação educacional de forma a orientar alunos na prevenção de doenças epidemiológicas transmissíveis, com ênfase na COVID-19. O principal artefato dessa abordagem é um sistema de recomendação via software Chatbot, denominado DinoBot. Esse software simula um diálogo entre indivíduos através da inteligência artificial e suas funcionalidades podem ser acessadas em diversos dispositivos. A análise e validação do DinoBot será por meio do modelo Technology Acceptance Model, com resultados que confirmam a eficácia do sistema desenvolvido.