Search

Data augmentation is a fundamental technique in machine learning that plays a crucial role in expanding the size of training datasets. By applying various transformations or modifications to existing data, data augmentation enhances the generalization and robustness of machine learning models. In recent years, the development of several libraries has simplified the utilization of diverse data augmentation strategies across different tasks. This paper focuses on the exploration of the most widely adopted libraries specifically designed for data augmentation in computer vision tasks. Here, we aim to provide a comprehensive survey of publicly available data augmentation libraries, facilitating practitioners to navigate these resources effectively. Through a curated taxonomy, we present an organized classification of the different approaches employed by these libraries, along with accompanying application examples. By examining the techniques of each library, practitioners can make informed decisions in selecting the most suitable augmentation techniques for their computer vision projects. To ensure the accessibility of this valuable information, a dedicated public website named DALib has been created. This website serves as a centralized repository where the taxonomy, methods, and examples associated with the surveyed data augmentation libraries can be explored. By offering this comprehensive resource, we aim to empower practitioners and contribute to the advancement of computer vision research and applications through effective utilization of data augmentation techniques.

We present IVL-SYNTHSFM-v2, a dataset of 4000 images depicting five objects from different viewpoints and under different lighting conditions and acquisition setups. The images have been rendered from 3D scenes with varying camera positions, lights, camera depth of field, and motion blur. Images depict one of five objects each acquired using eight different acquisition setups (scenes). 100 images have been rendered from each scene. The dataset is intended to be used to evaluate 3D reconstruction algorithms. The varying imaging conditions are introduced to challenge the algorithms to address realistic, non-ideal situations. The dataset provides pixel-precise ground truth to perform accurate evaluations. We demonstrate the usefulness of IVL-SYNTHSFM-v2 by assessing state-of-the-art 3D reconstruction algorithms.

La presentazione e comunicazione delle caratteristiche di prodotti, così come di raccolte di oggetti o servizi, sono di fondamentale importanza in diversi campi. Il caso d’uso più evidente è dato dalla necessità di produttori e negozi di pubblicizzare le proprie offerte ai potenziali clienti: a tal scopo vengono frequentemente impiegati cataloghi cartacei o digitali. Inoltre, i cataloghi possono avere obiettivi diversi dalla pubblicizzazione di prodotti fisici, possono anche presentare della conoscenza (ad esempio in una biblioteca o in un museo) oppure dei servizi forniti da una azienda o da un individuo. Con la crescente diffusione di smartphone, tablet ed head-mounted display si aprono nuove possibilità per i cataloghi, che possono essere presentati sfruttando i concetti della Mixed Reality (MR), i quali consentono l’inserimento di elementi virtuali all'interno del mondo reale. A tal fine, questa tesi propone un framework computazionale che supporta la creazione di cataloghi MR attraverso le fasi di progettazione, sviluppo e fruizione del catalogo stesso. La creazione di cataloghi MR coinvolge diversi aspetti; in questa ricerca vengono indagati quelli chiave, inclusi la ricostruzione tridimensionale della geometria, l'acquisizione dell'aspetto del materiale, la presentazione e fruizione del catalogo MR. I ricercatori hanno ampiamente studiato questi campi separatamente; questa tesi si propone quindi di affrontare problemi specifici del loro utilizzo nella creazione di un catalogo MR. La ricostruzione 3D della geometria trova impiego nella creazione di modelli tridimensionali virtuali di elementi reali da visualizzare nei cataloghi MR: viene quindi affrontato il problema della selezione del metodo di ricostruzione 3D più adatto. Oltre alla geometria, una corretta riproduzione dell’aspetto degli elementi virtuali è fondamentale per farli apparire realistici. Questa tesi propone la progettazione e lo sviluppo di un dispositivo portatile ed economico per l'acq, Presentation and communication of characteristics about products and, in general, collections of objects or services are of crucial relevance in several fields. The most prominent application is given by the necessity of manufacturers and stores to advertise their offers to potential customers. To this purpose, paper or digital catalogs are often employed. Furthermore, catalogs may have goals that differ from the advertisement of physical products. They may also build on the presentation of knowledge (e.g. in a library or museum) or refer to services provided by some company or individual. With the growing diffusion of smartphones, tablets, and head-mounted displays, new possibilities arise for the catalogs to be presented by taking advantage of Mixed Reality (MR) concepts that allow the blending of virtual elements within the real world. To this end, this thesis proposes a computational framework that supports the creation of MR catalogs throughout the stages of design, development, and fruition of the catalog itself. The creation of MR catalogs involves several aspects. The key ones among them are investigated in this research. These include 3D geometry reconstruction, material appearance acquisition, and presentation of the MR catalog. Researchers have largely investigated these fields separately. This thesis aims to tackle specific problems of their usage in the creation of an MR catalog. 3D geometry reconstruction finds use in the creation of virtual 3D models of real elements for display in MR catalogs. The problem of selecting the most suitable 3D reconstruction method is addressed. In addition to 3D geometry, the look and feel of virtual elements are crucial to make them appear realistic. This thesis proposes the design and development of a low cost portable device for material appearance acquisition. Finally, the presentation of the MR catalog involves the rendering of the virtual elements and interaction with the user. In order to accomplish that, a generi

The availability of high-resolution data and accurate ground truth is essential to evaluate and compare methods and algorithms properly. Moreover, it is often difficult to acquire real data for a given application domain that is sufficiently representative and heterogeneous in terms of scene representation, acquisition conditions, point of view, etc. To overcome the limitations of available datasets, this paper presents a new synthetic, multi-purpose dataset called ENRICH for testing photogrammetric and computer vision algorithms. Compared to existing datasets, ENRICH offers higher resolution images rendered with different lighting conditions, camera orientations, scales, and fields of view. Specifically, ENRICH is composed of three sub-datasets: ENRICH-Aerial, ENRICH-Square, and ENRICH-Statue, each exhibiting different characteristics. We show the usefulness of the proposed dataset on several examples of photogrammetry and computer vision-related tasks such as: evaluation of hand-crafted and deep learning-based local features, effects of ground control points (GCPs) configuration on the 3D accuracy, and monocular depth estimation. We make ENRICH publicly available at: https://github.com/davidemarelli/ENRICH.

We present SfM Flow, a Blender add-on that provides a toolset for the evaluation of three-dimensional reconstructions obtained form images. 3D reconstruction is increasingly becoming popular and, to date, many techniques are available to perform it. Choosing which technology to use for a specific 3D reconstruction task can be resource and time-consuming. By using this tool it is possible to create images of a virtual 3D scene, perform 3D reconstructions starting from the generated images using Structure from Motion pipelines, and evaluate the accuracy of the obtained 3D reconstruction. The evaluation is carried out comparing the 3D reconstruction with the virtual scene's geometry, that constitutes an exact ground truth, without the need for complex setup and dedicated hardware for the acquisition of a real scene. Furthermore, SfM Flow includes support for different lighting, depth of field, and motion blur setups, thus allowing to stress the 3D reconstruction pipelines under common critical conditions.

In the last few years, Augmented Reality, Virtual Reality, and Artificial Intelligence (AI) have been increasingly employed in different application domains. Among them, the retail market presents the opportunity to allow people to check the appearance of accessories, makeup, hairstyle, hair color, and clothes on themselves, exploiting virtual try-on applications. In this paper, we propose an eyewear virtual try-on experience based on a framework that leverages advanced deep learning-based computer vision techniques. The virtual try-on is performed on a 3D face reconstructed from a single input image. In designing our system, we started by studying the underlying architecture, components, and their interactions. Then, we assessed and compared existing face reconstruction approaches. To this end, we performed an extensive analysis and experiments for evaluating their design, complexity, geometry reconstruction errors, and reconstructed texture quality. The experiments allowed us to select the most suitable approach for our proposed try-on framework. Our system considers actual glasses and face sizes to provide a realistic fit estimation using a markerless approach. The user interacts with the system by using a web application optimized for desktop and mobile devices. Finally, we performed a usability study that showed an above-average score of our eyewear virtual try-on application.

Smart mirrors are devices that can display any kind of information and can interact with the user using touch and voice commands. Different kinds of smart mirrors exist: general purpose, medical, fashion, and other task specific ones. General purpose smart mirrors are suitable for home environments but the exiting ones offer similar, limited functionalities. In this paper, we present a general-purpose smart mirror that integrates several functionalities, standard and advanced, to support users in their everyday life. Among the advanced functionalities are the capabilities of detecting a person’s emotions, the short- and long-term monitoring and analysis of the emotions, a double authentication protocol to preserve the privacy, and the integration of Alexa Skills to extend the applications of the smart mirrors. We exploit a deep learning technique to develop most of the smart functionalities. The effectiveness of the device is demonstrated by the performances of the implemented functionalities, and the evaluation in terms of its usability with real users.

Virtual try-on allows people to check the appearance of accessories, makeup, hairstyle, hair color, clothes, and potentially more on themselves. In this paper, we propose an eyewear virtual try-on experience that is performed on a 3D face reconstructed from an input image allowing the user to see the virtual face and eyeglasses from different viewpoints. The try-on process takes into account real face and glasses sizes to provide a realistic fit estimation; it is fully automated and only requires a face picture and selection of eyeglasses frames to test.

This article presents a dataset with 4000 synthetic images portraying five 3D models from different viewpoints under varying lighting conditions. Depth of field and motion blur have also been used to generate realistic images. For each object, 8 scenes with different combinations of lighting, depth of field and motion blur are created and images are taken from 100 points of view. Data also includes information about camera intrinsic and extrinsic calibration parameters for each image as well as the ground truth geometry of the 3D models. The images were rendered using Blender. The aim of this dataset is to allow evaluation and comparison of different solutions for 3D reconstruction of objects starting from a set of images taken under different realistic acquisition setups.

Virtual try-on is a technology that allows people to virtually check the appearance of accessories, makeup, hairstyle, hair color, clothes and potentially more on themselves. The virtual try-on presents many advantages over real try-on, it speeds up the process providing the possibility to test hundreds of products without the need to reach a physical store. In this paper we propose a virtual try-on web application specific for eyeglasses and sunglasses that can be easily used by simply taking a picture of a face and selecting the desired frames. The try-on process is performed on a 3D face reconstructed from the input image allowing the user to see the virtual face and glasses from different viewpoints. The try-on process is fully automated and does not require the user to provide anything else but the picture and selection of the glasses frames to test.

Structure from Motion (SfM) is a pipeline that allows three-dimensional reconstruction starting from a collection of images. A typical SfM pipeline comprises different processing steps each of which tackles a different problem in the reconstruction pipeline. Each step can exploit different algorithms to solve the problem at hand and thus many different SfM pipelines can be built. How to choose the SfM pipeline best suited for a given task is an important question. In this paper we report a comparison of different state-of-the-art SfM pipelines in terms of their ability to reconstruct different scenes. We also propose an evaluation procedure that stresses the SfM pipelines using real dataset acquired with high-end devices as well as realistic synthetic dataset. To this end, we created a plug-in module for the Blender software to support the creation of synthetic datasets and the evaluation of the SfM pipeline. The use of synthetic data allows us to easily have arbitrarily large and diverse datasets with, in theory, infinitely precise ground truth. Our evaluation procedure considers both the reconstruction errors as well as the estimation errors of the camera poses used in the reconstruction.

Structure from Motion (SfM) is a pipeline that allows three-dimensional reconstruction starting from a collection of images. A typical SfM pipeline comprises different processing steps each of which tackle a different problem in the reconstruction pipeline. Each step can exploit different algorithms to solve the problem at hand and thus many different SfM pipelines can be built. There are many SfM pipelines available in the literature. How to choose the best among them? We present a Blender plug-in that provides an easy to use tool to compare them under different conditions using both real and synthetic datasets.

In this work we consider systems where the measurements and control signals are transmitted over networks that are affected by independent and identically distributed (i.i.d.) packet losses. In this setting, we study two separate problems. The first one is the design of an offline state estimator. The second one is the design of a linear quadratic regulator (LQR). Both designs take the statistics of the network and the system into account, and are optimal in the sense of minimizing given cost functions. It turns out that these two designs, of which each has an associated Riccati equation, are dual. We show that the convergence of the Riccati equation associated to this dual system formulation is a necessary and sufficient condition for stability of the estimator and controller. Finally, we compare the performance of the proposed offline estimator with those of other estimators available from the literature.