Cookies
O website necessita de alguns cookies e outros recursos semelhantes para funcionar. Caso o permita, o INESC TEC irá utilizar cookies para recolher dados sobre as suas visitas, contribuindo, assim, para estatísticas agregadas que permitem melhorar o nosso serviço. Ver mais
Aceitar Rejeitar
  • Menu
Publicações

2024

Community detection in interval-weighted networks

Autores
Alves, H; Brito, P; Campos, P;

Publicação
Data Min. Knowl. Discov.

Abstract

2024

Enhancing Forest Fire Detection and Monitoring Through Satellite Image Recognition: A Comparative Analysis of Classification Algorithms Using Sentinel-2 Data

Autores
Brito, T; Pereira, I; Costa, P; Lima, J;

Publicação
Communications in Computer and Information Science

Abstract
Worldwide, forests have been harassed by fire in recent years. Either by human intervention or other reasons, the history of the burned area is increasing considerably, harming fauna and flora. It is essential to detect an early ignition for fire-fighting authorities can act quickly, decreasing the impact of forest damage impacts. The proposed system aims to improve nature monitoring and improve the existing surveillance systems through satellite image recognition. The soil recognition via satellite images can determine the sensor modules’ best position and provide crucial input information for artificial intelligence-based systems. For this, satellite images from the Sentinel-2 program are used to generate forest density maps as updated as possible. Four classification algorithms make the Tree Cover Density (TCD) map, consisting of the Gaussian Mixture Model (GMM), Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbors (K-NN), which identify zones by training known regions. The results demonstrate a comparison between the algorithms through their performance in recognizing the forest, grass, pavement, and water areas by Sentinel-2 images. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

2024

Enhancing Object Detection in Maritime Environments Using Metadata

Autores
Fernandes, DS; Bispo, J; Bento, LC; Figueiredo, M;

Publicação
PROGRESS IN PATTERN RECOGNITION, IMAGE ANALYSIS, COMPUTER VISION, AND APPLICATIONS, CIARP 2023, PT II

Abstract
Over the years, many solutions have been suggested in order to improve object detection in maritime environments. However, none of these approaches uses flight information, such as altitude, camera angle, time of the day, and atmospheric conditions, to improve detection accuracy and network robustness, even though this information is often available and captured by the UAV. This work aims to develop a network unaffected by image-capturing conditions, such as altitude and angle. To achieve this, metadata was integrated into the neural network, and an adversarial learning training approach was employed. This was built on top of the YOLOv7, which is a state-of-the-art realtime object detector. To evaluate the effectiveness of this methodology, comprehensive experiments and analyses were conducted. Findings reveal that the improvements achieved by this approach are minimal when trying to create networks that generalize more across these specific domains. The YOLOv7 mosaic augmentation was identified as one potential responsible for this minimal impact because it also enhances the model's ability to become invariant to these image-capturing conditions. Another potential cause is the fact that the domains considered (altitude and angle) are not orthogonal with respect to their impact on captured images. Further experiments should be conducted using datasets that offer more diverse metadata, such as adverse weather and sea conditions, which may be more representative of real maritime surveillance conditions. The source code of this work is publicly available at https://git hub.com/ipleiria-robotics/maritime-metadata-adaptation.

2024

Image Transfer over MQTT in IoT: Message Segmentation and Encryption for Remote Indicator Panels

Autores
Valente, D; Brito, T; Correia, M; Carvalho, JA; Lima, J;

Publicação
Communications in Computer and Information Science - Optimization, Learning Algorithms and Applications

Abstract

2024

Sample Size Analysis for a Production Line Study of Time

Autores
da Silva, MI; Vaz, CB;

Publicação
Lecture Notes in Mechanical Engineering

Abstract
Setting labor standards is an important topic to operational and strategic planning which requires the time studies establishment. This paper applies the statistical method for the definition of a sample size in order to define a reliable cycle time for a real industrial process. For the case study it is considered a welding process performed by a single operator that does the load and unload of components in 4 different welding machines. In order to perform the time studies, it is necessary to collect continuously data in the production line by measuring the time taken for the operator to perform the task. In order to facilitate the measurements, the task is divided into small elements with visible start and end points, called Measurement Points, in which the measurement process is applied. Afterwards, the statistical method enables to determine the sample size of observations to calculate the reliable cycle time. For the welding process presented, it is stated that the sample size defined through the statistical method is 20. Thus, these time observations of the task are continuously collected in order to obtain a reliable cycle time for this welding process. This time study can be implemented in similar way in other industrial processes. © 2024, The Author(s), under exclusive license to Springer Nature Switzerland AG.

2024

Condition Invariance for Autonomous Driving by Adversarial Learning

Autores
Silva, DTE; Cruz, RPM;

Publicação
PROGRESS IN PATTERN RECOGNITION, IMAGE ANALYSIS, COMPUTER VISION, AND APPLICATIONS, CIARP 2023, PT I

Abstract
Object detection is a crucial task in autonomous driving, where domain shift between the training and the test set is one of the main reasons behind the poor performance of a detector when deployed. Some erroneous priors may be learned from the training set, therefore a model must be invariant to conditions that might promote such priors. To tackle this problem, we propose an adversarial learning framework consisting of an encoder, an object-detector, and a condition-classifier. The encoder is trained to deceive the condition-classifier and aid the object-detector as much as possible throughout the learning stage, in order to obtain highly discriminative features. Experiments showed that this framework is not very competitive regarding the trade-off between precision and recall, but it does improve the ability of the model to detect smaller objects and some object classes.

  • 1
  • 3741