Sperimentazione di Deep Metric Loss per Self-Supervised Information Retrieval Systems su CORD19.

Dopo lo sviluppo dei primi casi di Covid-19 in Cina nell’autunno del 2019, ad inizio 2020 l’intero pianeta è precipitato in una pandemia globale che ha stravolto le nostre vite con conseguenze che non si vivevano dall’influenza spagnola. La grandissima quantità di paper scientifici in continua pubblicazione sul coronavirus e virus ad esso affini ha portato alla creazione di un unico dataset dinamico chiamato CORD19 e distribuito gratuitamente. Poter reperire informazioni utili in questa mole di dati ha ulteriormente acceso i riflettori sugli information retrieval systems, capaci di recuperare in maniera rapida ed efficace informazioni preziose rispetto a una domanda dell'utente detta query. Di particolare rilievo è stata la TREC-COVID Challenge, competizione per lo sviluppo di un sistema di IR addestrato e testato sul dataset CORD19. Il problema principale è dato dal fatto che la grande mole di documenti è totalmente non etichettata e risulta dunque impossibile addestrare modelli di reti neurali direttamente su di essi. Per aggirare il problema abbiamo messo a punto nuove soluzioni self-supervised, a cui abbiamo applicato lo stato dell'arte del deep metric learning e dell'NLP. Il deep metric learning, che sta avendo un enorme successo soprattuto nella computer vision, addestra il modello ad "avvicinare" tra loro immagini simili e "allontanare" immagini differenti. Dato che sia le immagini che il testo vengono rappresentati attraverso vettori di numeri reali (embeddings) si possano utilizzare le stesse tecniche per "avvicinare" tra loro elementi testuali pertinenti (e.g. una query e un paragrafo) e "allontanare" elementi non pertinenti. Abbiamo dunque addestrato un modello SciBERT con varie loss, che ad oggi rappresentano lo stato dell'arte del deep metric learning, in maniera completamente self-supervised direttamente e unicamente sul dataset CORD19, valutandolo poi sul set formale TREC-COVID attraverso un sistema di IR e ottenendo risultati interessanti.

Deep learning models for the automatic analysis of volleyball plays

Collecting and analysing data is an important element in any field of human activity and research. Even in sports, collecting and analyzing statistical data is attracting a growing interest. Some exemplar use cases are: improvement of technical/tactical aspects for team coaches, definition of game strategies based on the opposite team play or evaluation of the performance of players. Other advantages are related to taking more precise and impartial judgment in referee decisions: a wrong decision can change the outcomes of important matches. Finally, it can be useful to provide better representations and graphic effects that make the game more engaging for the audience during the match. Nowadays it is possible to delegate this type of task to automatic software systems that can use cameras or even hardware sensors to collect images or data and process them. One of the most efficient methods to collect data is to process the video images of the sporting event through mixed techniques concerning machine learning applied to computer vision. As in other domains in which computer vision can be applied, the main tasks in sports are related to object detection, player tracking, and to the pose estimation of athletes. The goal of the present thesis is to apply different models of CNNs to analyze volleyball matches. Starting from video frames of a volleyball match, we reproduce a bird's eye view of the playing court where all the players are projected, reporting also for each player the type of action she/he is performing.

Automatic detection of the inferior alveolar nerve in CT scans with CenterNet

The inferior alveolar nerve (IAN) lies within the mandibular canal, named inferior alveolar canal in literature. The detection of this nerve is important during maxillofacial surgeries or for creating dental implants. The poor quality of cone-beam computed tomography (CBCT) and computed tomography (CT) scans and/or bone gaps within the mandible increase the difficulty of this task, posing a challenge to human experts who are going to manually detect it and resulting in a time-consuming task.Therefore this thesis investigates two methods to automatically detect the IAN: a non-data driven technique and a deep-learning method. The latter tracks the IAN position at each frame leveraging detections obtained with the deep neural network CenterNet, fined-tuned for our task, and temporal and spatial information.

Concepts and methods for efficient decentralized learning in federated settings

Deep Neural Networks (DNNs) have revolutionized a wide range of applications beyond traditional machine learning and artificial intelligence fields, e.g., computer vision, healthcare, natural language processing and others. At the same time, edge devices have become central in our society, generating an unprecedented amount of data which could be used to train data-hungry models such as DNNs. However, the potentially sensitive or confidential nature of gathered data poses privacy concerns when storing and processing them in centralized locations. To this purpose, decentralized learning decouples model training from the need of directly accessing raw data, by alternating on-device training and periodic communications. The ability of distilling knowledge from decentralized data, however, comes at the cost of facing more challenging learning settings, such as coping with heterogeneous hardware and network connectivity, statistical diversity of data, and ensuring verifiable privacy guarantees. This Thesis proposes an extensive overview of decentralized learning literature, including a novel taxonomy and a detailed description of the most relevant system-level contributions in the related literature for privacy, communication efficiency, data and system heterogeneity, and poisoning defense. Next, this Thesis presents the design of an original solution to tackle communication efficiency and system heterogeneity, and empirically evaluates it on federated settings. For communication efficiency, an original method, specifically designed for Convolutional Neural Networks, is also described and evaluated against the state-of-the-art. Furthermore, this Thesis provides an in-depth review of recently proposed methods to tackle the performance degradation introduced by data heterogeneity, followed by empirical evaluations on challenging data distributions, highlighting strengths and possible weaknesses of the considered solutions. Finally, this Thesis presents a novel perspective on the usage of Knowledge Distillation as a mean for optimizing decentralized learning systems in settings characterized by data heterogeneity or system heterogeneity. Our vision on relevant future research directions close the manuscript.

Computational methods applied to undeciphered scripts from the Aegean and Cyprus: Linear A and Cypro-Minoan

The study of ancient, undeciphered scripts presents unique challenges, that depend both on the nature of the problem and on the peculiarities of each writing system. In this thesis, I present two computational approaches that are tailored to two different tasks and writing systems. The first of these methods is aimed at the decipherment of the Linear A afraction signs, in order to discover their numerical values. This is achieved with a combination of constraint programming, ad-hoc metrics and paleographic considerations. The second main contribution of this thesis regards the creation of an unsupervised deep learning model which uses drawings of signs from ancient writing system to learn to distinguish different graphemes in the vector space. This system, which is based on techniques used in the field of computer vision, is adapted to the study of ancient writing systems by incorporating information about sequences in the model, mirroring what is often done in natural language processing. In order to develop this model, the Cypriot Greek Syllabary is used as a target, since this is a deciphered writing system. Finally, this unsupervised model is adapted to the undeciphered Cypro-Minoan and it is used to answer open questions about this script. In particular, by reconstructing multiple allographs that are not agreed upon by paleographers, it supports the idea that Cypro-Minoan is a single script and not a collection of three script like it was proposed in the literature. These results on two different tasks shows that computational methods can be applied to undeciphered scripts, despite the relatively low amount of available data, paving the way for further advancement in paleography using these methods.

Mind the gaps: cognitive-inspired AI for high-level visual sensemaking. Towards abstract concept image classification

The abundance of visual data and the push for robust AI are driving the need for automated visual sensemaking. Computer Vision (CV) faces growing demand for models that can discern not only what images "represent," but also what they "evoke." This is a demand for tools mimicking human perception at a high semantic level, categorizing images based on concepts like freedom, danger, or safety. However, automating this process is challenging due to entropy, scarcity, subjectivity, and ethical considerations. These challenges not only impact performance but also underscore the critical need for interoperability. This dissertation focuses on abstract concept-based (AC) image classification, guided by three technical principles: situated grounding, performance enhancement, and interpretability. We introduce ART-stract, a novel dataset of cultural images annotated with ACs, serving as the foundation for a series of experiments across four key domains: assessing the effectiveness of the end-to-end DL paradigm, exploring cognitive-inspired semantic intermediaries, incorporating cultural and commonsense aspects, and neuro-symbolic integration of sensory-perceptual data with cognitive-based knowledge. Our results demonstrate that integrating CV approaches with semantic technologies yields methods that surpass the current state of the art in AC image classification, outperforming the end-to-end deep vision paradigm. The results emphasize the role semantic technologies can play in developing both effective and interpretable systems, through the capturing, situating, and reasoning over knowledge related to visual data. Furthermore, this dissertation explores the complex interplay between technical and socio-technical factors. By merging technical expertise with an understanding of human and societal aspects, we advocate for responsible labeling and training practices in visual media. These insights and techniques not only advance efforts in CV and explainable artificial intelligence but also propel us toward an era of AI development that harmonizes technical prowess with deep awareness of its human and societal implications.

Reti Neurali Ricorrenti per la generazione di musica simbolica

Il Deep Learning ha radicalmente trasformato il mondo del Machine Learning migliorando lo stato dell'arte in diversi campi che spaziano dalla computer vision al natural language processing. Non fermandosi a problemi di classificazione, negli ultimi anni, applicazioni di tipo generativo hanno portato alla creazione di immagini realistiche e documenti letterali. Il mondo della musica non è esente da una moltitudine di esperimenti nello stesso campo, con risultati ancora acerbi ma comunque potenzialmente interessanti. In questa tesi verrà discussa l'applicazione di un di modello appartenente alla famiglia del Deep Learning per la generazione di musica simbolica.

Production Lot Detection through Synthetic Generated Data

Correctness of information gathered in production environments is an essential part of quality assurance processes in many industries, this task is often performed by human resources who visually take annotations in various steps of the production flow. Depending on the performed task the correlation between where exactly the information is gathered and what it represents is more than often lost in the process. The lack of labeled data places a great boundary on the application of deep neural networks aimed at object detection tasks, moreover supervised training of deep models requires a great amount of data to be available. Reaching an adequate large collection of labeled images through classic techniques of data annotations is an exhausting and costly task to perform, not always suitable for every scenario. A possible solution is to generate synthetic data that replicates the real one and use it to fine-tune a deep neural network trained on one or more source domains to a different target domain. The purpose of this thesis is to show a real case scenario where the provided data were both in great scarcity and missing the required annotations. Sequentially a possible approach is presented where synthetic data has been generated to address those issues while standing as a training base of deep neural networks for object detection, capable of working on images taken in production-like environments. Lastly, it compares performance on different types of synthetic data and convolutional neural networks used as backbones for the model.

Tecniche di deep learning per il monitoraggio del processo nell’industria dello pneumatico

Il seguente elaborato affronta l'implementazione di un algoritmo che affronta un problema di controllo di processo in ambito industriale utilizzando algoritmi di object detection. Infatti, il progetto concordato con il professore Di Stefano si è svolto in collaborazione con l’azienda Pirelli, nell’ambito della produzione di pneumatici. Lo scopo dell'algoritmo implementato è di verificare il preciso orientamento di elementi grafici della copertura, utilizzati dalle case automobilistiche per equipaggiare correttamente le vetture. In particolare, si devono individuare delle scritte sul battistrada della copertura e identificarne la posizione rispetto ad altri elementi fissati su di essa. La tesi affronta questo task in due parti distinte: la prima consiste nel training di algoritmi di deep learning per il riconoscimento degli elementi grafici e del battistrada, la seconda è un decisore che opera a valle del primo sistema utilizzando gli output delle reti allenate.

Metodologie di Corrispondenza Stereo Basate su Deep Learning per Superfici Altamente Riflettenti e Trasparenti: Dataset e Architettura

Nell’ambito della Stereo Vision, settore della Computer Vision, partendo da coppie di immagini RGB, si cerca di ricostruire la profondità della scena. La maggior parte degli algoritmi utilizzati per questo compito ipotizzano che tutte le superfici presenti nella scena siano lambertiane. Quando sono presenti superfici non lambertiane (riflettenti o trasparenti), gli algoritmi stereo esistenti sbagliano la predizione della profondità. Per risolvere questo problema, durante l’esperienza di tirocinio, si è realizzato un dataset contenente oggetti trasparenti e riflettenti che sono la base per l’allenamento della rete. Agli oggetti presenti nelle scene sono associate annotazioni 3D usate per allenare la rete. Invece, nel seguente lavoro di tesi, utilizzando l’algoritmo RAFT-Stereo [1], rete allo stato dell’arte per la stereo vision, si analizza come la rete modifica le sue prestazioni (predizione della disparità) se al suo interno viene inserito un modulo per la segmentazione semantica degli oggetti. Si introduce questo layer aggiuntivo perché, trovare la corrispondenza tra due punti appartenenti a superfici lambertiane, risulta essere molto complesso per una normale rete. Si vuole utilizzare l’informazione semantica per riconoscere questi tipi di superfici e così migliorarne la disparità. È stata scelta questa architettura neurale in quanto, durante l’esperienza di tirocinio riguardante la creazione del dataset Booster [2], è risultata la migliore su questo dataset. L’obiettivo ultimo di questo lavoro è vedere se il riconoscimento di superfici non lambertiane, da parte del modulo semantico, influenza la predizione della disparità migliorandola. Nell’ambito della stereo vision, gli elementi riflettenti e trasparenti risultano estremamente complessi da analizzare, ma restano tuttora oggetto di studio dati gli svariati settori di applicazione come la guida autonoma e la robotica.

Dataset Fusion application on deep learning Object Detection task

Application of dataset fusion techniques to an object detection task, involving the use of deep learning as convolutional neural networks, to manage to create a single RCNN architecture able to inference with good performances on two distinct datasets with different domains.

An Embedded In-Cabin Lightweight High-Performance 3D Gaze Estimation System

Gaze estimation has gained interest in recent years for being an important cue to obtain information about the internal cognitive state of humans. Regardless of whether it is the 3D gaze vector or the point of gaze (PoG), gaze estimation has been applied in various fields, such as: human robot interaction, augmented reality, medicine, aviation and automotive. In the latter field, as part of Advanced Driver-Assistance Systems (ADAS), it allows the development of cutting-edge systems capable of mitigating road accidents by monitoring driver distraction. Gaze estimation can be also used to enhance the driving experience, for instance, autonomous driving. It also can improve comfort with augmented reality components capable of being commanded by the driver's eyes. Although, several high-performance real-time inference works already exist, just a few are capable of working with only a RGB camera on computationally constrained devices, such as a microcontroller. This work aims to develop a low-cost, efficient and high-performance embedded system capable of estimating the driver's gaze using deep learning and a RGB camera. The proposed system has achieved near-SOTA performances with about 90% less memory footprint. The capabilities to generalize in unseen environments have been evaluated through a live demonstration, where high performance and near real-time inference were obtained using a webcam and a Raspberry Pi4.

Depth Estimation in Stereo Biomedical Images via Proxy-Supervised Deep Learning

In order to estimate depth through supervised deep learning-based stereo methods, it is necessary to have access to precise ground truth depth data. While the gathering of precise labels is commonly tackled by deploying depth sensors, this is not always a viable solution. For instance, in many applications in the biomedical domain, the choice of sensors capable of sensing depth at small distances with high precision on difficult surfaces (that present non-Lambertian properties) is very limited. It is therefore necessary to find alternative techniques to gather ground truth data without having to rely on external sensors. In this thesis, two different approaches have been tested to produce supervision data for biomedical images. The first aims to obtain input stereo image pairs and disparities through simulation in a virtual environment, while the second relies on a non-learned disparity estimation algorithm in order to produce noisy disparities, which are then filtered by means of hand-crafted confidence measures to create noisy labels for a subset of pixels. Among the two, the second approach, which is referred in literature as proxy-labeling, has shown the best results and has even outperformed the non-learned disparity estimation algorithm used for supervision.

Identificazione di caratteristiche rilevanti in capi di abbigliamento mediante deep-learning

Il mondo della moda è in continua e costante evoluzione, non solo dal punto di vista sociale, ma anche da quello tecnologico. Nel corso del presente elaborato si è studiata la possibilità di riconoscere e segmentare abiti presenti in una immagine utilizzando reti neurali profonde e approcci moderni. Sono state, quindi, analizzate reti quali FasterRCNN, MaskRCNN, YOLOv5, FashionPedia e Match-RCNN. In seguito si è approfondito l’addestramento delle reti neurali profonde in scenari di alta parallelizzazione e su macchine dotate di molteplici GPU al fine di ridurre i tempi di addestramento. Inoltre si è sperimentata la possibilità di creare una rete per prevedere se un determinato abito possa avere successo in futuro analizzando semplicemente dati passati e una immagine del vestito in questione. Necessaria per tali compiti è stata, inoltre, una approfondita analisi dei dataset esistenti nel mondo della moda e dei metodi per utilizzarli per l’addestramento. Il presente elaborato è stato svolto nell’ambito del progetto FA.RE.TRA. per il quale l'Università di Bologna svolge un compito di consulenza per lo studio di fattibilità su reti neurali in grado di svolgere i compiti menzionati.

Convolutional Neural Network Architectures for Template Matching

The Neural Networks customized and tested in this thesis (WaldoNet, FlowNet and PatchNet) are a first exploration and approach to the Template Matching task. The possibilities of extension are therefore many and some are proposed below. During my thesis, I have analyzed the functioning of the classical algorithms and adapted with deep learning algorithms. The features extracted from both the template and the query images resemble the keypoints of the SIFT algorithm. Then, instead of similarity function or keypoints matching, WaldoNet and PatchNet use the convolutional layer to compare the features, while FlowNet uses the correlational layer. In addition, I have identified the major challenges of the Template Matching task (affine/non-affine transformations, intensity changes...) and solved them with a careful design of the dataset.