Technological innovation, conservation and enhancement are key elements for promoting cultural heritage and attracting visitors worldwide. Cultural tourism represents a significant economic and social strategy, capable of stimulating regional development and revitalizing marginalized areas. In recent years, artificial intelligence (AI) has transformed the sector, offering new ways to engage with and preserve cultural assets. This study explores the application of advanced AI techniques, particularly Convolutional Neural Networks (CNNs), to improve the classification and recognition of images related to architectural heritage. A deep learning algorithm specifically designed for cultural heritage enhancement is presented, focusing on the automatic classification of images of buildings and monuments. The analysis includes a comparison between pre-trained models and custom models, highlighting the performance of different approaches. The experimental results demonstrate that our ensemble model achieved 90% accuracy in classifying architectural heritage elements across 10 categories, with the Majority Vote Ensemble approach outperforming individual models by 6-10%. This improved classification accuracy enables more reliable automated systems for cultural heritage documentation and interactive tourist experiences. The research addresses key challenges in architectural heritage classification including variations in preservation state, lighting conditions, and complex backgrounds with multiple elements. The study also investigates the impact of data augmentation and class balancing techniques on model performance, demonstrating how these methods can mitigate limitations in training data availability. The developed ensemble combines state-of-the-art CNN architectures like ResNet50 and EfficientNet with custom models, leveraging their complementary strengths to achieve robust classification across diverse architectural styles and conditions. The adopted approach is supervised, using a training dataset where images are pre-labeled according to specific categories. This allows the algorithm to learn and subsequently predict the categories of new images based on the acquired information. A practical application is proposed to improve visitor experiences and heritage management, paving the way for future technological advancements in the field.

Advances in artificial intelligence are revolutionizing agricultural diagnostics, particularly in addressing the critical challenge of cassava disease detection. Cassava, a vital food source for millions worldwide, faces significant yield losses due to various diseases that threaten food security in developing regions. This research presents a novel approach integrating transfer learning with explainable AI to create a robust disease detection system. Through extensive experimentation with multiple deep learning architectures, our ResNet-based model achieves a remarkable accuracy of 92% in distinguishing among four major cassava diseases and healthy specimens. The integration of SHAP (SHapley Additive exPlanations) technology provides unprecedented transparency in the model’s decision-making process, allowing stakeholders to understand how the neural network identifies disease-specific features. Our system demonstrates particular strength in identifying Cassava Mosaic Disease, achieving 98% accuracy, while maintaining robust performance across bacterial blight, brown spot and green mite detection. The methodology presented here not only advances the technical frontier of agricultural AI but also provides a practical tool for enhancing food security through early disease detection. This research establishes a foundation for developing accessible and interpretable AI systems that can be deployed in resource-limited agricultural settings, potentially transforming how farmers manage crop health in the digital age.

The growing disparity between maize crop demand and actual production is concerning for both the food industry and farmers. Worldwide production of 1147.7 million MT of maize is insufficient to meet the demand of approximately 1149.96 million MT. Diseases like Turcicum Leaf Blight and Rust significantly hamper maize production. Manual disease detection, classification, severity calculation, and estimating crop loss are time-consuming and demand specific expertise. Hence, there’s a pressing need for automatic disease detection, severity prediction, and crop loss estimation. Machine learning and deep learning techniques, known for their success in pattern recognition and data analysis, have encouraged researchers to apply them in detecting diseases and estimating crop losses in maize. While existing literature showcases potential in disease detection, there’s a lack of reliable, real-world labeled datasets for training these models. Also, the focus on severity prediction and crop loss estimation is lacking in previous works. The paper provides a comprehensive overview of deep-learning approaches for Crop Loss Estimation in Maize Agriculture.

The rapidly increasing demand for energy and the consequent depletion of non-renewable energy sources pose significant challenges. Seeking alternatives, renewable sources like solar cells come into focus. Nevertheless, their limited efficiency hinders practical application and motivates researchers to develop more efficient solar cells. Through an examination of effectiveness, design viability, and fabrication costs, Dye-Sensitized Solar Cells (DSSC) emerge as superior to other photovoltaic solar cells. In particular the dye component is crucial for how well the DSSC works as it absorbs light from the sun.The paper investigates the topic related to forecasting the absorption maxima (λmax) of dyes and presents an overview of the proposals in the literature that use neural networks to forecast it. In addition, it discusses the main challenges related to this relevant topic, evidencing the need to address these challenges.

Agriculture is a vital industry for both the world’s food supply and economic health. The increasing global population and the demand for sustainable food production have led to the emergence of Artificial Intelligence (AI) as a game-changing technology to tackle agricultural issues. In this article, we explore the opportunities and challenges of AI-driven agriculture and provide an overview of the most promising applications and related ethical and practical issues.

Diabetic Retinopathy (DR) is a complication of diabetes, caused by a damage to the blood vessels in the light-sensitive tissue of the retina. Since it affects the eyes, it can determine visual impairment or even blindness. Considering the number of diabetic patients worldwide, it is clear that effective screening of potential DR patients is of utmost importance. While direct and indirect ophthalmoscopy are the main methods for evaluating DR, artificial intelligence is on the rise in vision care. DR is detectable by analyzing data from patients’ fundus photographs, and is therefore a disease that artificial intelligence tools can effectively support. In this paper, we present some preliminary numerical results obtained in discriminating between eye fundi of healthy individuals and of people with severe diabetic retinopathy, by using a Multiple Instance Learning approach.

At the end of 2019, the World Health Organization (WHO) referred that the Public Health Commission of Hubei Province, China, reported cases of severe and unknown pneumonia, characterized by fever, malaise, dry cough, dyspnoea and respiratory failure, which occurred in the urban area of Wuhan. A new coronavirus, SARS-CoV-2, was identified as responsible for the lung infection, now called COVID-19 (coronavirus disease 2019). Since then there has been an exponential growth of infections and at the beginning of March 2020 the WHO declared the epidemic a global emergency. An early diagnosis of those carrying the virus becomes crucial to contain the spread, morbidity and mortality of the pandemic. The definitive diagnosis is made through specific tests, among which imaging tests play an important role in the care path of the patient with suspected or confirmed COVID-19. Patients with serious COVID-19 typically experience viral pneumonia.

Among skin cancers, melanoma is the most aggressive and most lethal form. Despite these terrible premises, an excision treatment carried out thanks to an early diagnosis is almost always decisive, guaranteeing the patient's survival. The early detection of melanoma is hampered by the extreme similarity of melanoma with other skin lesions such as dysplastic nevi. The current research is aimed at defining software solutions that support the computerized diagnosis of lesions for the detection of melanoma. To date, the proposals, both in terms of algorithms and frameworks, have focused on the dichotomous distinction of melanoma from benign lesions. However, the current debate on Dysplastic Nevi Syndrome (DNS), makes issues relating to the nature of the lesions, central to subjects who present a large number of moles throughout the body. In fact, individuals with DNS have a greater chance of being attacked by melanoma. The classification task relating to the distinction of dysplastic nevi from common ones is totally unexplored. In this document, we consider the difficult task of applying multiple-instance learning (MIL) approaches to discriminate melanoma from dysplastic nevi and outline an even more complex challenge related to the classification of dysplastic nevi from common ones. In particular, we introduce the application of a MIL approach that uses spherical separation surfaces. Since the results seem promising, we conclude that a MIL technique could be the basis of more sophisticated tools useful for detecting skin lesions.

Artificial Intelligent Systems are increasingly used to support early diagnosis of multiple relevant diseases. The spread of these systems is boosted by the application of machine learning techniques on datasets (also in the form of videos and images) obtained from different information sources. A key role is played by artificial vision systems that are in charge of reasoning on data acquired from different devices, including smartphones. The facility to disseminate and share information let to the globalization of medical protocols previously used just in some world's areas. This is the case of tongue inspection, widely used in Traditional Chinese Medicine (TCM) to perform a diagnosis, which allows physicians to obtain useful indications on the state of internal organs by observing the color and the consistency of patient's tongue. The current interest in tongue's image analysis is also motivated by the possibility of performing a first self-analysis on a possible disease suggesting further medical investigation. The paper is a non-exhaustive overview of the features most frequently used in artificial vision systems contextualized to tongue analysis. It highlights shortcomings in some of the existing studies and provides insights for future research. Our work aims to provide a unifying view that can support the researchers working on Tongue Colored Image Analysis.

Correct interpretation of images may be crucial for early disease detection. A growing number of medical instruments are image-oriented and produce a large quantity of image data, typically in the DICOM format, which contain spatio-temporal features together with alpha-numeric information regarding patients. Dealing with this high-dimensional datasets is a complex and time-consuming task. In addition the diffusion of smartphones and tablets requires the development of technological features enabling the medical team to check on helthcare processes on-the-go and freely access and send image and data for case analysis and collaborative diagnostic. This paper presents SIMPATICO 3D (Sistema Informativo Medico PATologIe COmplesse) a system supporting scientists and physicians by providing facilities for case studies analysis and diagnostic imaging in a shared virtual environment and details the features of SIMPATICO 3D Mobile (standing for Evolution Imaging System 3D for Mobile), that extends SIMPATICO 3D with dedicated functions for the mobile environment.

After a brief survey on well established methods for image classification, we focus on a recently proposed Multiple Istance Learning (MIL) method which is suitable for applications in image processing.