Umberto Michelucci

Francesca Venturini, Michela Sperti, Umberto Michelucci et al.

The olive oil sector produces a substantial impact in the Mediterranean's economy and lifestyle. Many studies exist which try to optimize the different steps in the olive oil's production process. One of the main challenges for olive oil producers is the ability to asses and control the quality during the production cycle. For this purpose, several parameters need to be determined, such as the acidity, the UV absorption or the ethyl esters content. To achieve this, samples must be sent to an approved laboratory for chemical analysis. This approach is expensive and cannot be performed very frequently, making quality control of olive oil a real challenge. This work explores a new approach based on fluorescence spectroscopy and artificial intelligence (namely, 1-D convolutional neural networks) to predict the five chemical quality indicators of olive oil (acidity, peroxide value, UV spectroscopic parameters $K_{270}$ and $K_{232}$, and ethyl esters) from simple fluorescence spectra. Fluorescence spectroscopy is a very attractive optical technique since it does not require sample preparation, is non destructive, and, as shown in this work, can be easily implemented in small and cost-effective sensors. The results indicate that the proposed approach gives exceptional results in the quality determination and would make the continuous quality control of olive oil during and after the production process a reality. Additionally, this novel methodology presents potential applications as a support for quality specifications of olive oil, as defined by the European regulation.

Francesca Venturini, Michela Sperti, Umberto Michelucci et al.

This dataset encompasses fluorescence spectra and chemical parameters of 24 olive oil samples from the 2019-2020 harvest provided by the producer Conde de Benalua, Granada, Spain. The oils are characterized by different qualities: 10 extra virgin olive oil (EVOO), 8 virgin olive oil (VOO), and 6 lampante olive oil (LOO) samples. For each sample, the dataset includes fluorescence spectra obtained with two excitation wavelengths, oil quality, and five chemical parameters necessary for the quality assessment of olive oil. The fluorescence spectra were obtained by exciting the samples at 365 nm and 395 nm under identical conditions. The dataset includes the values of the following chemical parameters for each olive oil sample: acidity, peroxide value, K270, K232, ethyl esters, and the quality of the samples (EVOO, VOO, or LOO). The dataset offers a unique possibility for researchers in food technology to develop machine learning models based on fluorescence data for the quality assessment of olive oil due to the availability of both spectroscopic and chemical data. The dataset can be used, for example, to predict one or multiple chemical parameters or to classify samples based on their quality from fluorescence spectra.

Francesca Venturini, Silvan Fluri, Manas Mejari et al.

This work systematically investigates the oxidation of extra virgin olive oil (EVOO) under accelerated storage conditions with UV absorption and total fluorescence spectroscopy. With the large amount of data collected, it proposes a method to monitor the oil's quality based on machine learning applied to highly-aggregated data. EVOO is a high-quality vegetable oil that has earned worldwide reputation for its numerous health benefits and excellent taste. Despite its outstanding quality, EVOO degrades over time owing to oxidation, which can affect both its health qualities and flavour. Therefore, it is highly relevant to quantify the effects of oxidation on EVOO and develop methods to assess it that can be easily implemented under field conditions, rather than in specialized laboratories. The following study demonstrates that fluorescence spectroscopy has the capability to monitor the effect of oxidation and assess the quality of EVOO, even when the data are highly aggregated. It shows that complex laboratory equipment is not necessary to exploit fluorescence spectroscopy using the proposed method and that cost-effective solutions, which can be used in-field by non-scientists, could provide an easily-accessible assessment of the quality of EVOO.

Umberto Michelucci, Francesca Venturini

The application of machine learning to physics problems is widely found in the scientific literature. Both regression and classification problems are addressed by a large array of techniques that involve learning algorithms. Unfortunately, the measurement errors of the data used to train machine learning models are almost always neglected. This leads to estimations of the performance of the models (and thus their generalisation power) that is too optimistic since it is always assumed that the target variables (what one wants to predict) are correct. In physics, this is a dramatic deficiency as it can lead to the belief that theories or patterns exist where, in reality, they do not. This paper addresses this deficiency by deriving formulas for commonly used metrics (both for regression and classification problems) that take into account measurement errors of target variables. The new formulas give an estimation of the metrics which is always more pessimistic than what is obtained with the classical ones, not taking into account measurement errors. The formulas given here are of general validity, completely model-independent, and can be applied without limitations. Thus, with statistical confidence, one can analyze the existence of relationships when dealing with measurements with errors of any kind. The formulas have wide applicability outside physics and can be used in all problems where measurement errors are relevant to the conclusions of studies.

Safouane El Ghazouali, Youssef Mhirit, Ali Oukhrid et al.

In the realm of computer vision, the integration of advanced techniques into the processing of RGB-D camera inputs poses a significant challenge, given the inherent complexities arising from diverse environmental conditions and varying object appearances. Therefore, this paper introduces FusionVision, an exhaustive pipeline adapted for the robust 3D segmentation of objects in RGB-D imagery. Traditional computer vision systems face limitations in simultaneously capturing precise object boundaries and achieving high-precision object detection on depth map as they are mainly proposed for RGB cameras. To address this challenge, FusionVision adopts an integrated approach by merging state-of-the-art object detection techniques, with advanced instance segmentation methods. The integration of these components enables a holistic (unified analysis of information obtained from both color \textit{RGB} and depth \textit{D} channels) interpretation of RGB-D data, facilitating the extraction of comprehensive and accurate object information. The proposed FusionVision pipeline employs YOLO for identifying objects within the RGB image domain. Subsequently, FastSAM, an innovative semantic segmentation model, is applied to delineate object boundaries, yielding refined segmentation masks. The synergy between these components and their integration into 3D scene understanding ensures a cohesive fusion of object detection and segmentation, enhancing overall precision in 3D object segmentation. The code and pre-trained models are publicly available at https://github.com/safouaneelg/FusionVision/.

Safouane El Ghazouali, Arnaud Gucciardi, Francesca Venturini et al.

Object detection in remotely sensed satellite pictures is fundamental in many fields such as biophysical, and environmental monitoring. While deep learning algorithms are constantly evolving, they have been mostly implemented and tested on popular ground-based taken photos. This paper critically evaluates and compares a suite of advanced object detection algorithms customized for the task of identifying aircraft within satellite imagery. Using the large HRPlanesV2 dataset, together with a rigorous validation with the GDIT dataset, this research encompasses an array of methodologies including YOLO versions 5 and 8, Faster RCNN, CenterNet, RetinaNet, RTMDet, and DETR, all trained from scratch. This exhaustive training and validation study reveal YOLOv5 as the preeminent model for the specific case of identifying airplanes from remote sensing data, showcasing high precision and adaptability across diverse imaging conditions. This research highlight the nuanced performance landscapes of these algorithms, with YOLOv5 emerging as a robust solution for aerial object detection, underlining its importance through superior mean average precision, Recall, and Intersection over Union scores. The findings described here underscore the fundamental role of algorithm selection aligned with the specific demands of satellite imagery analysis and extend a comprehensive framework to evaluate model efficacy. The benchmark toolkit and codes, available via https://github.com/toelt-llc/FlightScope_Bench, aims to further exploration and innovation in the realm of remote sensing object detection, paving the way for improved analytical methodologies in satellite imagery applications.

Safouane El Ghazouali, Nicola Venturi, Michael Rueegsegger et al.

Recent advances in deep learning for remote sensing rely heavily on large annotated datasets, yet acquiring high-quality ground truth for geometric, radiometric, and multi-domain tasks remains costly and often infeasible. In particular, the lack of accurate depth annotations, controlled illumination variations, and multi-scale paired imagery limits progress in monocular depth estimation, domain adaptation, and super-resolution for aerial scenes. We present SyMTRS, a large-scale synthetic dataset generated using a high-fidelity urban simulation pipeline. The dataset provides high-resolution RGB aerial imagery (2048 x 2048), pixel-perfect depth maps, night-time counterparts for domain adaptation, and aligned low-resolution variants for super-resolution at x2, x4, and x8 scales. Unlike existing remote sensing datasets that focus on a single task or modality, SyMTRS is designed as a unified multi-task benchmark enabling joint research in geometric understanding, cross-domain robustness, and resolution enhancement. We describe the dataset generation process, its statistical properties, and its positioning relative to existing benchmarks. SyMTRS aims to bridge critical gaps in remote sensing research by enabling controlled experiments with perfect geometric ground truth and consistent multi-domain supervision. The results obtained in this work can be reproduced from this Github repository: https://github.com/safouaneelg/SyMTRS.

Morteza Kiani Haftlang, Mohammadhossein Malmir, Foroutan Parand et al.

Medical image segmentation is a critical task in clinical workflows, particularly for the detection and delineation of pathological regions. While convolutional architectures like U-Net have become standard for such tasks, their limited receptive field restricts global context modeling. Recent efforts integrating transformers have addressed this, but often result in deep, computationally expensive models unsuitable for real-time use. In this work, we present a novel end-to-end lightweight architecture designed specifically for real-time binary medical image segmentation. Our model combines a Swin Transformer-like encoder with a U-Net-like decoder, connected via skip pathways to preserve spatial detail while capturing contextual information. Unlike existing designs such as Swin Transformer or U-Net, our architecture is significantly shallower and competitively efficient. To improve the encoder's ability to learn meaningful features without relying on large amounts of labeled data, we first train it using Barlow Twins, a self-supervised learning method that helps the model focus on important patterns by reducing unnecessary repetition in the learned features. After this pretraining, we fine-tune the entire model for our specific task. Experiments on benchmark binary segmentation tasks demonstrate that our model achieves competitive accuracy with substantially reduced parameter count and faster inference, positioning it as a practical alternative for deployment in real-time and resource-limited clinical environments. The code for our method is available at Github repository: https://github.com/mkianih/Barlow-Swin.

Safouane El Ghazouali, Umberto Michelucci

AI models rely on annotated data to learn pattern and perform prediction. Annotation is usually a labor-intensive step that require associating labels ranging from a simple classification label to more complex tasks such as object detection, oriented bounding box estimation, and instance segmentation. Traditional tools often require extensive manual input, limiting scalability for large datasets. To address this, we introduce VisioFirm, an open-source web application designed to streamline image labeling through AI-assisted automation. VisioFirm integrates state-of-the-art foundation models into an interface with a filtering pipeline to reduce human-in-the-loop efforts. This hybrid approach employs CLIP combined with pre-trained detectors like Ultralytics models for common classes and zero-shot models such as Grounding DINO for custom labels, generating initial annotations with low-confidence thresholding to maximize recall. Through this framework, when tested on COCO-type of classes, initial prediction have been proven to be mostly correct though the users can refine these via interactive tools supporting bounding boxes, oriented bounding boxes, and polygons. Additionally, VisioFirm has on-the-fly segmentation powered by Segment Anything accelerated through WebGPU for browser-side efficiency. The tool supports multiple export formats (YOLO, COCO, Pascal VOC, CSV) and operates offline after model caching, enhancing accessibility. VisioFirm demonstrates up to 90\% reduction in manual effort through benchmarks on diverse datasets, while maintaining high annotation accuracy via clustering of connected CLIP-based disambiguate components and IoU-graph for redundant detection suppression. VisioFirm can be accessed from \href{https://github.com/OschAI/VisioFirm}{https://github.com/OschAI/VisioFirm}.

Safouane El Ghazouali, Arnaud Gucciardi, Umberto Michelucci

Self-rewarding have emerged recently as a powerful tool in the field of Natural Language Processing (NLP), allowing language models to generate high-quality relevant responses by providing their own rewards during training. This innovative technique addresses the limitations of other methods that rely on human preferences. In this paper, we build upon the concept of self-rewarding models and introduce its vision equivalent for Text-to-Image generative AI models. This approach works by fine-tuning diffusion model on a self-generated self-judged dataset, making the fine-tuning more automated and with better data quality. The proposed mechanism makes use of other pre-trained models such as vocabulary based-object detection, image captioning and is conditioned by the a set of object for which the user might need to improve generated data quality. The approach has been implemented, fine-tuned and evaluated on stable diffusion and has led to a performance that has been evaluated to be at least 60\% better than existing commercial and research Text-to-image models. Additionally, the built self-rewarding mechanism allowed a fully automated generation of images, while increasing the visual quality of the generated images and also more efficient following of prompt instructions. The code used in this work is freely available on https://github.com/safouaneelg/SRT2I.

Umberto Michelucci

Training neural networks means solving a high-dimensional optimization problem. Normally the goal is to minimize a loss function that depends on what is called the network function, or in other words the function that gives the network output given a certain input. This function depends on a large number of parameters, also known as weights, that depends on the network architecture. In general the goal of this optimization problem is to find the global minimum of the network function. In this paper it is discussed how due to how neural networks are designed, the neural network function present a very large symmetry in the parameter space. This work shows how the neural network function has a number of equivalent minima, in other words minima that give the same value for the loss function and the same exact output, that grows factorially with the number of neurons in each layer for feed forward neural network or with the number of filters in a convolutional neural networks. When the number of neurons and layers is large, the number of equivalent minima grows extremely fast. This will have of course consequences for the study of how neural networks converges to minima during training. This results is known, but in this paper for the first time a proper mathematical discussion is presented and an estimate of the number of equivalent minima is derived.

Umberto Michelucci, Francesca Venturini

Machine learning (ML) models have achieved strikingly high accuracies in spectroscopic classification tasks, often without a clear proof that those models used chemically meaningful features. Existing studies have linked these results to data preprocessing choices, noise sensitivity, and model complexity, but no unifying explanation is available so far. In this work, we show that these phenomena arise naturally from the intrinsic high dimensionality of spectral data. Using a theoretical analysis grounded in the Feldman-Hajek theorem and the concentration of measure, we show that even infinitesimal distributional differences, caused by noise, normalisation, or instrumental artefacts, may become perfectly separable in high-dimensional spaces. Through a series of specific experiments on synthetic and real fluorescence spectra, we illustrate how models can achieve near-perfect accuracy even when chemical distinctions are absent, and why feature-importance maps may highlight spectrally irrelevant regions. We provide a rigorous theoretical framework, confirm the effect experimentally, and conclude with practical recommendations for building and interpreting ML models in spectroscopy.

Arnaud Gucciardi, Safouane El Ghazouali, Francesca Venturini et al.

This paper presents an annotated dataset of brain MRI images designed to advance the field of brain symmetry study. Magnetic resonance imaging (MRI) has gained interest in analyzing brain symmetry in neonatal infants, and challenges remain due to the vast size differences between fetal and adult brains. Classification methods for brain structural MRI use scales and visual cues to assess hemisphere symmetry, which can help diagnose neonatal patients by comparing hemispheres and anatomical regions of interest in the brain. Using the Developing Human Connectome Project dataset, this work presents a dataset comprising cerebral images extracted as slices across selected portions of interest for clinical evaluation . All the extracted images are annotated with the brain's midline. All the extracted images are annotated with the brain's midline. From the assumption that a decrease in symmetry is directly related to possible clinical pathologies, the dataset can contribute to a more precise diagnosis because it can be used to train deep learning model application in neonatal cerebral MRI anomaly detection from postnatal infant scans thanks to computer vision. Such models learn to identify and classify anomalies by identifying potential asymmetrical patterns in medical MRI images. Furthermore, this dataset can contribute to the research and development of methods using the relative symmetry of the two brain hemispheres for crucial diagnosis and treatment planning.

Nathalie Alexander, Arnaud Gucciardi, Umberto Michelucci

Accurate segmentation of infant brain MRI is essential for quantifying developmental changes in structure and complexity. However, ongoing myelination and reduced tissue contrast make automated segmentation particularly challenging. This study systematically compared segmentation accuracy and its impact on volumetric and fractal dimension (FD) estimates in infant brain MRI using the Baby Open Brains (BOB) dataset (71 scans, 1-9 months). Two methods, SynthSeg and SamSeg, were evaluated against expert annotations using Dice, Intersection over Union, 95th-percentile Hausdorff distance, and Normalised Mutual Information. SynthSeg outperformed SamSeg across all quality metrics (mean Dice > 0.8 for major regions) and provided volumetric estimates closely matching the manual reference (mean +4% [-28% - 71%]). SamSeg systematically overestimated ventricular and whole-brain volumes (mean +76% [-12% - 190%]). Segmentation accuracy improved with age, consistent with increasing tissue contrast during myelination. Fractal dimension a(FD) nalyses revealed significant regional differences between SynthSeg and expert segmentations, and Bland-Altman limits of agreement indicated that segmentation-related FD variability exceeded most group differences reported in developmental cohorts. Volume and FD deviations were positively correlated across structures, indicating that segmentation bias directly affects FD estimation. Overall, SynthSeg provided the most reliable volumetric and FD results for paediatric MRI, yet small morphological differences in volume and FD should be interpreted with caution due to segmentation-related uncertainty.

Lorenzo Monti, Tatiana Muraveva, Brian Sheridan et al.

In data-driven scientific discovery, a challenge lies in classifying well-characterized phenomena while identifying novel anomalies. Current semi-supervised clustering algorithms do not always fully address this duality, often assuming that supervisory signals are globally representative. Consequently, methods often enforce rigid constraints that suppress unanticipated patterns or require a pre-specified number of clusters, rendering them ineffective for genuine novelty detection. To bridge this gap, we introduce CLiMB (CLustering in Multiphase Boundaries), a domain-informed framework decoupling the exploitation of prior knowledge from the exploration of unknown structures. Using a sequential two-phase approach, CLiMB first anchors known clusters using constrained partitioning, and subsequently applies density-based clustering to residual data to reveal arbitrary topologies. We demonstrate this framework on RR Lyrae stars data from the Gaia Data Release 3. CLiMB attains an Adjusted Rand Index of 0.829 with 90% seed coverage in recovering known Milky Way substructures, drastically outperforming heuristic and constraint-based baselines, which stagnate below 0.20. Furthermore, sensitivity analysis confirms CLiMB's superior data efficiency, showing monotonic improvement as knowledge increases. Finally, the framework successfully isolates three dynamical features (Shiva, Shakti, and the Galactic Disk) in the unlabelled field, validating its potential for scientific discovery.

Umberto Michelucci, Francesca Venturini

Machine learning (ML) is increasingly adopted in scientific research, yet the quality and reliability of results often depend on how experiments are designed and documented. Poor baselines, inconsistent preprocessing, or insufficient validation can lead to misleading conclusions about model performance. This paper presents a practical and structured guide for conducting ML experiments in scientific applications, focussing on reproducibility, fair comparison, and transparent reporting. We outline a step-by-step workflow, from dataset preparation to model selection and evaluation, and propose metrics that account for overfitting and instability across validation folds, including the Logarithmic Overfitting Ratio (LOR) and the Composite Overfitting Score (COS). Through recommended practices and example reporting formats, this work aims to support researchers in establishing robust baselines and drawing valid evidence-based insights from ML models applied to scientific problems.

Umberto Michelucci, Francesca Venturini

Machine learning is vital in high-stakes domains, yet conventional validation methods rely on averaging metrics like mean squared error (MSE) or mean absolute error (MAE), which fail to quantify extreme errors. Worst-case prediction failures can have substantial consequences, but current frameworks lack statistical foundations for assessing their probability. In this work a new statistical framework, based on Extreme Value Theory (EVT), is presented that provides a rigorous approach to estimating worst-case failures. Applying EVT to synthetic and real-world datasets, this method is shown to enable robust estimation of catastrophic failure probabilities, overcoming the fundamental limitations of standard cross-validation. This work establishes EVT as a fundamental tool for assessing model reliability, ensuring safer AI deployment in new technologies where uncertainty quantification is central to decision-making or scientific analysis.

Umberto Michelucci, Francesca Venturini

Fluorescence spectroscopy is a fundamental tool in life sciences and chemistry, widely used for applications such as environmental monitoring, food quality control, and biomedical diagnostics. However, analysis of spectroscopic data with deep learning, in particular of fluorescence excitation-emission matrices (EEMs), presents significant challenges due to the typically small and sparse datasets available. Furthermore, the analysis of EEMs is difficult due to their high dimensionality and overlapping spectral features. This study proposes a new approach that exploits domain adaptation with pretrained vision models, alongside a novel interpretability algorithm to address these challenges. Thanks to specialised feature engineering of the neural networks described in this work, we are now able to provide deeper insights into the physico-chemical processes underlying the data. The proposed approach is demonstrated through the analysis of the oxidation process in extra virgin olive oil (EVOO) during ageing, showing its effectiveness in predicting quality indicators and identifying the spectral bands, and thus the molecules involved in the process. This work describes a significantly innovative approach in the use of deep learning for spectroscopy, transforming it from a black box into a tool for understanding complex biological and chemical processes.

Umberto Michelucci

In this article, we will look at autoencoders. This article covers the mathematics and the fundamental concepts of autoencoders. We will discuss what they are, what the limitations are, the typical use cases, and we will look at some examples. We will start with a general introduction to autoencoders, and we will discuss the role of the activation function in the output layer and the loss function. We will then discuss what the reconstruction error is. Finally, we will look at typical applications as dimensionality reduction, classification, denoising, and anomaly detection. This paper contains the notes of a PhD-level lecture on autoencoders given in 2021.

Umberto Michelucci, Michela Sperti, Dario Piga et al.

This paper presents the intrinsic limit determination algorithm (ILD Algorithm), a novel technique to determine the best possible performance, measured in terms of the AUC (area under the ROC curve) and accuracy, that can be obtained from a specific dataset in a binary classification problem with categorical features {\sl regardless} of the model used. This limit, namely the Bayes error, is completely independent of any model used and describes an intrinsic property of the dataset. The ILD algorithm thus provides important information regarding the prediction limits of any binary classification algorithm when applied to the considered dataset. In this paper the algorithm is described in detail, its entire mathematical framework is presented and the pseudocode is given to facilitate its implementation. Finally, an example with a real dataset is given.

Francesca Venturini, Michela Sperti, Umberto Michelucci et al.

Extra virgin olive oil (EVOO) is the highest quality of olive oil and is characterized by highly beneficial nutritional properties. The large increase in both consumption and fraud, for example through adulteration, creates new challenges and an increasing demand for developing new quality assessment methodologies that are easier and cheaper to perform. As of today, the determination of olive oil quality is performed by producers through chemical analysis and organoleptic evaluation. The chemical analysis requires the advanced equipment and chemical knowledge of certified laboratories, and has therefore a limited accessibility. In this work a minimalist, portable and low-cost sensor is presented, which can perform olive oil quality assessment using fluorescence spectroscopy. The potential of the proposed technology is explored by analyzing several olive oils of different quality levels, EVOO, virgin olive oil (VOO), and lampante olive oil (LOO). The spectral data were analyzed using a large number of machine learning methods, including artificial neural networks. The analysis performed in this work demonstrates the possibility of performing classification of olive oil in the three mentioned classes with an accuracy of 100$\%$. These results confirm that this minimalist low-cost sensor has the potential of substituting expensive and complex chemical analysis.

Umberto Michelucci, Michael Baumgartner, Francesca Venturini

Luminescence-based sensors for measuring oxygen concentration are widely used both in industry and research due to the practical advantages and sensitivity of this type of sensing. The measuring principle is the luminescence quenching by oxygen molecules, which results in a change of the luminescence decay time and intensity. In the classical approach, this change is related to an oxygen concentration using the Stern-Volmer equation. This equation, which in most of the cases is non-linear, is parametrized through device-specific constants. Therefore, to determine these parameters every sensor needs to be precisely calibrated at one or more known concentrations. This work explores an entirely new artificial intelligence approach and demonstrates the feasibility of oxygen sensing through machine learning. The specifically developed neural network learns very efficiently to relate the input quantities to the oxygen concentration. The results show a mean deviation of the predicted from the measured concentration of 0.5 percent air, comparable to many commercial and low-cost sensors. Since the network was trained using synthetically generated data, the accuracy of the model predictions is limited by the ability of the generated data to describe the measured data, opening up future possibilities for significant improvement by using a large number of experimental measurements for training. The approach described in this work demonstrates the applicability of artificial intelligence to sensing of sensors.