Abstract
Public lighting plays a very important role for society's safety and quality of life. The identification of faults in public lighting is essential for the maintenance and prevention of safety. Traditionally, this task depends on human action, through checking during the day, representing expenditure and waste of energy. Automatic detection with deep learning is an innovative solution that can be explored for locating and identifying of this kind of problem. In this study, we present a first approach, composed of several steps, intending to obtain the segmentation of public lighting, using Seville (Spain) as case study. A dataset called NLight was created from a nighttime image taken by the JL1-3B satellite, and four U-Net and FPN architectures were trained with different backbones to segment part of the NLight. The U-Net with InceptionResNetv2 proved to be the model with the best performance, obtained 761 of 815, correct locations (93.4%). This model was used to predict the segmentation of the remaining dataset. This study provides the location of lamps so that we can identify patterns and possible lighting failures in the future.

Abstract
The grape moth is a common pest that affects grapevines by consuming both fruit and foliage, rendering grapes deformed and unsellable. Integrated pest management for the grape moth heavily relies on pheromone traps, which serve a crucial function by identifying and tracking adult moth populations. This information is then used to determine the most appropriate time and method for implementing other control techniques. This study aims to find the best method for detecting small insects. We evaluate the following recent YOLO models: v5, v6, v7, and v8 for detecting and counting grape moths in insect traps. The best performance was achieved by YOLOv8, with an average precision of 92.4% and a counting error of 8.1%.

Abstract
In this study, we evaluated the use of small pre-trained 3D Convolutional Neural Networks (CNN) on land use and land cover (LULC) slide-window-based classification. We pre-trained the small models in a dataset with origin in the Eurosat dataset and evaluated the benefits of the transfer-learning plus fine-tuning for four different regions using Sentinel-2 L1C imagery (bands of 10 and 20m of spatial resolution), comparing the results to pre-trained models and trained from scratch. The models achieved an F1 Score of between 0.69-0.80 without significative change when pre-training the model. However, for small datasets, pre-training the model improved the classification by up to 3%.

Abstract
In recent years, the number of crimes with weapons has grown on a large scale worldwide, mainly in locations where enforcement is lacking or possessing weapons is legal. It is necessary to combat this type of criminal activity to identify criminal behavior early and allow police and law enforcement agencies immediate action.Despite the human visual structure being highly evolved and able to process images quickly and accurately if an individual watches something very similar for a long time, there is a possibility of slowness and lack of attention. In addition, large surveillance systems with numerous equipment require a surveillance team, which increases the cost of operation. There are several solutions for automatic weapon detection based on computer vision; however, these have limited performance in challenging contexts.A systematic review of the current literature on deep learning-based weapon detection was conducted to identify the methods used, the main characteristics of the existing datasets, and the main problems in the area of automatic weapon detection. The most used models were the Faster R-CNN and the YOLO architecture. The use of realistic images and synthetic data showed improved performance. Several challenges were identified in weapon detection, such as poor lighting conditions and the difficulty of small weapon detection, the last being the most prominent. Finally, some future directions are outlined with a special focus on small weapon detection.

Abstract
Grape varieties play an important role in wine's production chain, its identification is crucial for controlling and regulating the production. Nowadays, two techniques are widely used, ampelography and molecular analysis. However, there are problems with both of them. In this scenario, Deep Learning classifiers emerged as a tool to automatically classify grape varieties. A problem with the classification of on-field acquired images is that there is a lot of information unrelated to the target classification. In this study, the use of segmentation before classification to remove such unrelated information was analyzed. We used two grape varieties identification datasets to fine-tune a pre-trained EfficientNetV2S. Our results showed that segmentation can slightly improve classification performance if only unrelated information is removed.

Abstract