Abstract

Abstract
Easily accessible through tabletop experiments, paraxial fluids of light are emerging as promising platforms for the simulation and exploration of quantumlike phenomena. In particular, the analogy builds on a formal equivalence between the governing model for a Bose-Einstein condensate under the mean-field approximation and the model of laser propagation inside nonlinear optical media under the paraxial approximation. Yet, the fact that the role of time is played by the propagation distance in the analog system imposes strong bounds on the range of accessible phenomena due to the limited length of the nonlinear medium. In this Letter, we present an experimental approach to solve this limitation in the form of a digital feedback loop, which consists of the reconstruction of the optical states at the end of the system followed by their subsequent reinjection exploiting wavefront shaping techniques. The results enclosed demonstrate the potential of this approach to access unprecedented dynamics, paving the way for the observation of novel phenomena in these systems.

Abstract
TadGAN is a recent algorithm with competitive performance on time series anomaly detection. The detection process of TadGAN works by comparing observed data with generated data. A challenge in anomaly detection is that there are anomalies which are not easy to detect by analyzing the original time series but have a clear effect on its higher-order characteristics. We propose Meta-TadGAN, an adaptation of TadGAN that analyzes meta-level representations of time series. That is, it analyzes a time series that represents the characteristics of the time series, rather than the original time series itself. Results on benchmark datasets as well as real-world data from fire detectors shows that the new method is competitive with TadGAN. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

Abstract
Context. The GRAVITY+ upgrade implies a complete renewal of its adaptive optics (AO) systems. Its complex design, featuring moving components between the deformable mirrors and the wavefront sensors, requires the monitoring and auto-calibrating of the lateral mis-registrations of the system while in operation. Aims. For preset and target acquisition, large lateral registration errors must be assessed in open loop to bring the system to a state where the AO loop closes. In closed loop, these errors must be monitored and corrected, without impacting the science. Methods. With respect to the first requirement, our method is perturbative, with two-dimensional modes intentionally applied to the system and correlated to a reference interaction matrix. For the second requirement, we applied a non-perturbative approach that searches for specific patterns in temporal correlations in the closed loop telemetry. This signal is produced by the noise propagation through the AO loop. Results. Our methods were validated through simulations and on the GRAVITY+ development bench. The first method robustly estimates the lateral mis-registrations, in a single fit and with a sub-subaperture resolution while in an open loop. The second method is not absolute, but it does successfully bring the system towards a negligible mis-registration error, with a limited turbulence bias. Both methods proved to robustly work on a system still under development and not fully characterised. Conclusions. Tested with Shack-Hartmann wavefront sensors, the proposed methods are versatile and easily adaptable to other AO instruments, such as the pyramid, which stands as a baseline for all future AO systems. The non-perturbative method, not relying on an interaction matrix model and being sparse in the Fourier domain, is particularly suitable to the next generation of AO systems for extremely large telescopes that will present an unprecedented level of complexity and numbers of actuators.

Abstract
Modern agriculture faces important challenges for feeding a fast-growing planet's population in a sustainable way. One of the most important challenges faced by agriculture is the increasing destruction caused by pests to important crops. It is very important to control and manage pests in order to reduce the losses they cause. However, pest detection and monitoring are very resources consuming tasks. The recent development of computer vision-based technology has made it possible to automatize pest detection efficiently. In Mediterranean olive groves, the olive fly (Bactrocera oleae Rossi) is considered the key-pest of the crop. This paper presents olive fly detection using the lightweight YOLO-based model for versions 7 and 8, respectively, YOLOv7-tiny and YOLOv8n. The proposed object detection models were trained, validated, and tested using two different image datasets collected in various locations of Portugal and Greece. The images are constituted by sticky yellow trap photos and by McPhail trap photos with olive fly exemplars. The performance of the models was evaluated using precision, recall, and mAP.95. The YOLOV7-tiny model best performance is 88.3% of precision, 85% of Recall, 90% of mAP.50, and 53% of mAP.95. The YOLOV8n model best performance is 85% of precision, 85% of Recall, 90% mAP.50, and 55% of mAP.50 YOLO8n model achieved worst results than YOLOv7-tiny for a dataset without negative images (images without olive fly exemplars). Aiming at installing an experimental prototype in the olive grove, the YOLOv8n model was implemented in a Ubuntu Server 23.04 Raspberry PI 3 microcomputer.

Abstract
Effective image and video annotation is a fundamental pillar in computer vision and artificial intelligence, crucial for the development of accurate machine learning models. Object tracking and image retrieval techniques are essential in this process, significantly improving the efficiency and accuracy of automatic annotation. This paper systematically investigates object tracking and image acquisition techniques. It explores how these technologies can collectively enhance the efficiency and accuracy of the annotation processes for image and video datasets. Object tracking is examined for its role in automating annotations by tracking objects across video sequences, while image retrieval is evaluated for its ability to suggest annotations for new images based on existing data. The review encompasses diverse methodologies, including advanced neural networks and machine learning techniques, highlighting their effectiveness in various contexts like medical analyses and urban monitoring. Despite notable advancements, challenges such as algorithm robustness and effective human-AI collaboration are identified. This review provides valuable insights into these technologies' current state and future potential in improving image annotation processes, even showing existing applications of these techniques and their full potential when combined.