Monitoring coastal marine wildlife is crucial for biodiversity conservation, environmental management, and sustainable utilization of tourism-related natural assets. Conducting in situ censuses and population studies in extensive and remote marine habitats often faces logistical constraints, necessitating the adoption of advanced technologies to enhance the efficiency and accuracy of monitoring efforts. This study investigates the utilization of aerial imagery and deep learning methodologies for the automated detection, classification, and enumeration of marine-coastal species. A comprehensive dataset of high-resolution images, captured by drones and aircrafts over southern elephant seal (Mirounga leonina) and South American sea lion (Otaria flavescens) colonies in the Valdés Peninsula, Patagonia, Argentina, was curated and annotated. Using this annotated dataset, a deep learning framework was developed and trained to identify and classify individual animals. The resulting model may help produce automated, accurate population metrics that support the analysis of ecological dynamics. The resulting model achieved F1 scores of between 0.7 and 0.9, depending on the type of individual. Among its contributions, this methodology provided essential insights into the impacts of emergent threats, such as the outbreak of the highly pathogenic avian influenza virus H5N1 during the 2023 austral spring season, which caused significant mortality in these species.

Recently, nuclear abnormalities in avian erythrocytes have been used as biomarkers of genotoxicity in several species. Anomalous shapes are usually detected in the nuclei by means of microscopy inspection. However, due to inter- and intra-observer variability, the classification of these blood cell abnormalities could be problematic for replicating research. Deep learning, as a powerful image analysis technique, can be used in this context to improve standardization in identifying the biological configurations of medical and veterinary importance. In this study, we present a standardized deep learning model for identifying and classifying abnormal shapes in erythrocyte nuclei in blood smears of the hemispheric and synanthropic Kelp Gull (Larus dominicanus). We trained three convolutional backbones (ResNet34 and ResNet50 architectures) to obtain models capable of detecting and classifying these abnormalities in blood cells. The analysis was performed at three discrimination levels of classification, with broad categories subdivided into increasingly specific subcategories (level 1: “normal”, “abnormal”, “other”; level 2: “normal”, “ENAs”, “micronucleus”, “other”; level 3: “normal”, “irregular”, “displaced”, “enucleated”, “micronucleus”, “other”). The results were more than adequate and very similar in levels 1 and 2 (F1-score 84.6% and 83.6%, and accuracy 83.9% and 82.6%). In level 3, performance was lower (F1-score 65.9% and accuracy 80.8%). It can be concluded that the level 2 analysis should be considered the most appropriate as it is more specific than level 1, with similar quality of performance. This method has proven to be a fast, efficient, and standardized approach that reduces the dependence on human supervision in the classification of nuclear abnormalities in avian erythrocytes, and can be adapted to be used in similar contexts with reduced effort.

Sound level modeling has emerged as an essential tool for predicting acoustic environments. We present the development and analysis of models using a dataset previously applied for sound exceedance level modeling in the contiguous United States. This dataset comprises acoustic exceedance levels measured in diverse locations including National Park Service sites and urban environments. We applied advanced python libraries to train Random Forest regression models to predict exceedance levels from 99 geospatial variables. In total, 3 general and 5 ancillary fully data-driven models (not modeling actual physics of sound propagation) were developed, and the particular performance and limitations of each model is discussed. Results show promising predictive power, with R2 between 0.54 and 0.91 and root mean squared error between 1.77 and 5.97 dB, where models incorporating more urban information performed better. These results highlight the strength of the models, with performance variability primarily attributed to the limited coverage of diverse natural and urban environments in the current dataset. Results are accessible via an interactive online dashboard, allowing users without machine learning expertise to analyze different aspects of the models. This platform supports broader accessibility, encouraging a wider audience to engage with outdoor sound level modeling and its applications.

We present the development of an automatic dental diagnosis system based on artificial intelligence. The goal is to accurately support clinical decision-making in dentistry based on pantomographic images. We describe the construction of a large dataset of pantomographic images with high-quality annotations, the development of an intelligent model for semantic image interpretation and generation of presumptive diagnoses, and an application for end users. The model demonstrates the ability to correctly identify each individual tooth, detect multiple conditions and common pathologies in dental pieces, in case they arise, and generate a presumptive diagnostic text based on the information extracted from the image. The application, in the form of a digital web-based negatoscope, represents a technological advancement in radiology in line with the digital transformation era, offering a versatile and efficient tool for the interactive visualization of radiographic images, especially designed for dental professionals.

Las imágenes satelitales suelen presentar limitaciones en su resolución espacial y, en muchos casos, altos costos de adquisición, lo que restringe su uso en aplicaciones como el monitoreo urbano, la gestión territorial y el estudio de fauna. Este trabajo propone un enfoque innovador que aprovecha imágenes aéreas de alta resolución para entrenar un modelo de super-resolución basado en Redes Generativas Adversarias. En particular, se adapta el modelo ESRGAN (Enhanced Super-Resolution Generative Adversarial Network), optimizando sus parámetros con el fin de mejorar su eficiencia computacional y reducir los tiempos de entrenamiento. El modelo entrenado con imágenes aéreas se evalúa posteriormente sobre recortes satelitales de baja resolución, analizando su desempeño en factores de escala x2 y x4 mediante métricas estructurales, perceptuales y cromáticas (SSIM-Y, MS-SSIM, LPIPS y CIEDE2000). Los resultados muestran mejoras visuales claras, con mayor nitidez, mejor definición de bordes y una recuperación coherente de estructuras urbanas y elementos del terreno. Cuantitativamente, la escala x2 alcanza los valores más altos, mientras que la escala x4 mantiene un rendimiento estable y útil para aplicaciones prácticas. Estos hallazgos demuestran la viabilidad de transferir la capacidad de super-resolución desde imágenes aéreas hacia imágenes satelitales, incluso bajo diferencias espectrales y geométricas entre dominios. En conjunto, este trabajo establece una base sólida para el desarrollo de modelos de super-resolución satelital de bajo costo y alto impacto, y abre futuras líneas de investigación orientadas a ampliar los datos de entrenamiento, incorporar técnicas de domain adaptation y explorar arquitecturas específicas para sensores satelitales.