Search

Despite the recent advances in computer vision research, estimating the 3D human pose from single RGB images remains a challenging task, as multiple 3D poses can correspond to the same 2D projection on the image. In this context, depth data could help to disambiguate the 2D information by providing additional constraints about the distance between objects in the scene and the camera. Unfortunately, the acquisition of accurate depth data is limited to indoor spaces and usually is tied to specific depth technologies and devices, thus limiting generalization capabilities. In this paper, we propose a method able to leverage the benefits of depth information without compromising its broader applicability and adaptability in a predominantly RGB-camera-centric landscape. Our approach consists of a heatmap-based 3D pose estimator that, leveraging the paradigm of Privileged Information, is able to hallucinate depth information from the RGB frames given at inference time. More precisely, depth information is used exclusively during training by enforcing our RGB-based hallucination network to learn similar features to a backbone pre-trained only on depth data. This approach proves to be effective even when dealing with limited and small datasets. Experimental results reveal that the paradigm of Privileged Information significantly enhances the model's performance, enabling efficient extraction of depth information by using only RGB images., Comment: ECCV 2024 Workshop T-CAP: TOWARDS A COMPLETE ANALYSIS OF PEOPLE: FINE-GRAINED UNDERSTANDING FOR REAL-WORLD APPLICATIONS

The three-dimensional representation of objects or scenes starting from a set of images has been a widely discussed topic for years and has gained additional attention after the diffusion of NeRF-based approaches. However, an underestimated prerequisite is the knowledge of camera poses or, more specifically, the estimation of the extrinsic calibration parameters. Although excellent general-purpose Structure-from-Motion methods are available as a pre-processing step, their computational load is high and they require a lot of frames to guarantee sufficient overlapping among the views. This paper introduces KRONC, a novel approach aimed at inferring view poses by leveraging prior knowledge about the object to reconstruct and its representation through semantic keypoints. With a focus on vehicle scenes, KRONC is able to estimate the position of the views as a solution to a light optimization problem targeting the convergence of keypoints' back-projections to a singular point. To validate the method, a specific dataset of real-world car scenes has been collected. Experiments confirm KRONC's ability to generate excellent estimates of camera poses starting from very coarse initialization. Results are comparable with Structure-from-Motion methods with huge savings in computation. Code and data will be made publicly available., Comment: Accepted at ECCVW

We use motivic methods to give a quick proof of Berthelot's conjecture stating that the push-forward map in rigid cohomology of the structural sheaf along a smooth and proper map has a canonical structure of overconvergent F-isocrystal on the base., Comment: 28 pages

We study rare events in the extreme value statistics of stochastic symmetric jump processes with power tails in the distributions of the jumps, using the big-jump principle. The principle states that in the presence of stochastic processes with power tails statistics, if at a certain time a physical quantity takes on a value much larger than its typical value, this large fluctuation is realised through a single macroscopic jump that exceeds the typical scale of the process by several orders of magnitude. In particular, our estimation focuses on the asymptotic behaviour of the tail of the probability distribution of maxima, a fundamental quantity in a wide class of stochastic models used in chemistry to estimate reaction thresholds, in climatology for earthquake risk assessment, in finance for portfolio management, and in ecology for the collective behaviour of species. We determine the analytical form of the probability distribution of rare events in the extreme value statistics of three jump processes with power tails; L\'evy flights, L\'evy walks and the L\'evy-Lorentz gas. For the L\'evy flights, we re-obtain through the big-jump approach recent analytical results, extending their validity. For the L\'evy-Lorentz gas we show that the topology of the disordered lattice along which the walker moves induces memory effects in its dynamics, which influences the extreme value statistics. Our results are confirmed by extensive numerical simulations., Comment: 9 pages, 6 figures

Human behaviour strongly influences the spread of infectious diseases: understanding the interplay between epidemic dynamics and adaptive behaviours is essential to improve response strategies to epidemics, with the goal of containing the epidemic while preserving a sufficient level of operativeness in the population. Through activity-driven temporal networks, we formulate a general framework which models a wide range of adaptive behaviours and mitigation strategies, observed in real populations. We analytically derive the conditions for a widespread diffusion of epidemics in the presence of arbitrary adaptive behaviours, highlighting the crucial role of correlations between agents behaviour in the infected and in the susceptible state. We focus on the effects of sick-leave, comparing the effectiveness of different strategies in reducing the impact of the epidemic and preserving the system operativeness. We show the critical relevance of heterogeneity in individual behavior: in homogeneous networks, all sick-leave strategies are equivalent and poorly effective, while in heterogeneous networks, strategies targeting the most vulnerable nodes are able to effectively mitigate the epidemic, also avoiding a deterioration in system activity and maintaining a low level of absenteeism. Interestingly, with targeted strategies both the minimum of population activity and the maximum of absenteeism anticipate the infection peak, which is effectively flattened and delayed, so that full operativeness is almost restored when the infection peak arrives. We also provide realistic estimates of the model parameters for influenza-like illness, thereby suggesting strategies for managing epidemics and absenteeism in realistic populations., Comment: 16 pages, 8 figures

Millions of people around the world are infected with mosquito-borne diseases each year. One of the most dangerous species is Aedes aegypti, the main vector of viruses such as dengue, yellow fever, chikungunya, and Zika, among others. Mosquito prevention and eradication campaigns are essential to avoid major public health consequences. In this respect, entomological surveillance is an important tool. At present, this traditional monitoring tool is executed manually and requires digital transformation to help authorities make better decisions, improve their planning efforts, speed up execution, and better manage available resources. Therefore, new technological tools based on proven techniques need to be designed and developed. However, such tools should also be cost-effective, autonomous, reliable, and easy to implement, and should be enabled by connectivity and multi-platform software applications. This paper presents the design, development, and testing of an innovative system named MosquIoT. It is based on traditional ovitraps with embedded Internet of Things (IoT) and Tiny Machine Learning (TinyML) technologies, which enable the detection and quantification of Ae. aegypti eggs. This innovative and promising solution may help dynamically understand the behavior of Ae. aegypti populations in cities, shifting from the current reactive entomological monitoring model to a proactive and predictive digital one.

We perform accurate numerical experiments with fully-connected (FC) one-hidden layer neural networks trained with a discretized Langevin dynamics on the MNIST and CIFAR10 datasets. Our goal is to empirically determine the regimes of validity of a recently-derived Bayesian effective action for shallow architectures in the proportional limit. We explore the predictive power of the theory as a function of the parameters (the temperature $T$, the magnitude of the Gaussian priors $\lambda_1$, $\lambda_0$, the size of the hidden layer $N_1$ and the size of the training set $P$) by comparing the experimental and predicted generalization error. The very good agreement between the effective theory and the experiments represents an indication that global rescaling of the infinite-width kernel is a main physical mechanism for kernel renormalization in FC Bayesian standard-scaled shallow networks.

Rare events in the first-passage distributions of jump processes are capable of triggering anomalous reactions or series of events. Estimating their probability is particularly important when the jump probabilities have broad-tailed distributions, and rare events are therefore not so rare. We formulate a general approach for estimating the contribution of fast rare events to the exit probabilities in the presence of fat tailed distributions. Using this approach, we study three jump processes that are used to model a wide class of phenomena ranging from biology to transport in disordered systems, ecology and finance: discrete time random-walks, L\'evy walks and the L\'evy-Lorentz gas. We determine the exact form of the scaling function for the probability distribution of fast rare events, in which the jump process exits from an interval in a very short time at a large distance opposite to the starting point. In particular, we show that events occurring on time scales orders of magnitude smaller than the typical time scale of the process can make a significant contribution to the exit probability. Our results are confirmed by extensive numerical simulations., Comment: 6 pages, 4 figures

Technologies to enable safe and effective collaboration and coexistence between humans and robots have gained significant importance in the last few years. A critical component useful for realizing this collaborative paradigm is the understanding of human and robot 3D poses using non-invasive systems. Therefore, in this paper, we propose a novel vision-based system leveraging depth data to accurately establish the 3D locations of skeleton joints. Specifically, we introduce the concept of Pose Nowcasting, denoting the capability of the proposed system to enhance its current pose estimation accuracy by jointly learning to forecast future poses. The experimental evaluation is conducted on two different datasets, providing accurate and real-time performance and confirming the validity of the proposed method on both the robotic and human scenarios.

Neural Radiance Fields (NeRFs) have gained widespread recognition as a highly effective technique for representing 3D reconstructions of objects and scenes derived from sets of images. Despite their efficiency, NeRF models can pose challenges in certain scenarios such as vehicle inspection, where the lack of sufficient data or the presence of challenging elements (e.g. reflections) strongly impact the accuracy of the reconstruction. To this aim, we introduce CarPatch, a novel synthetic benchmark of vehicles. In addition to a set of images annotated with their intrinsic and extrinsic camera parameters, the corresponding depth maps and semantic segmentation masks have been generated for each view. Global and part-based metrics have been defined and used to evaluate, compare, and better characterize some state-of-the-art techniques. The dataset is publicly released at https://aimagelab.ing.unimore.it/go/carpatch and can be used as an evaluation guide and as a baseline for future work on this challenging topic., Comment: Accepted at ICIAP2023

Feature learning, or the ability of deep neural networks to automatically learn relevant features from raw data, underlies their exceptional capability to solve complex tasks. However, feature learning seems to be realized in different ways in fully-connected (FC) or convolutional architectures (CNNs). Empirical evidence shows that FC neural networks in the infinite-width limit eventually outperform their finite-width counterparts. Since the kernel that describes infinite-width networks does not evolve during training, whatever form of feature learning occurs in deep FC architectures is not very helpful in improving generalization. On the other hand, state-of-the-art architectures with convolutional layers achieve optimal performances in the finite-width regime, suggesting that an effective form of feature learning emerges in this case. In this work, we present a simple theoretical framework that provides a rationale for these differences, in one hidden layer networks. First, we show that the generalization performance of a finite-width FC network can be obtained by an infinite-width network, with a suitable choice of the Gaussian priors. Second, we derive a finite-width effective action for an architecture with one convolutional hidden layer and compare it with the result available for FC networks. Remarkably, we identify a completely different form of kernel renormalization: whereas the kernel of the FC architecture is just globally renormalized by a single scalar parameter, the CNN kernel undergoes a local renormalization, meaning that the network can select the local components that will contribute to the final prediction in a data-dependent way. This finding highlights a simple mechanism for feature learning that can take place in overparametrized shallow CNNs, but not in shallow FC architectures or in locally connected neural networks without weight sharing., Comment: 22 pages, 5 figures, 2 tables. Comments are welcome

The ability to leverage heterogeneous robotic experience from different robots and tasks to quickly master novel skills and embodiments has the potential to transform robot learning. Inspired by recent advances in foundation models for vision and language, we propose a multi-embodiment, multi-task generalist agent for robotic manipulation. This agent, named RoboCat, is a visual goal-conditioned decision transformer capable of consuming action-labelled visual experience. This data spans a large repertoire of motor control skills from simulated and real robotic arms with varying sets of observations and actions. With RoboCat, we demonstrate the ability to generalise to new tasks and robots, both zero-shot as well as through adaptation using only 100-1000 examples for the target task. We also show how a trained model itself can be used to generate data for subsequent training iterations, thus providing a basic building block for an autonomous improvement loop. We investigate the agent's capabilities, with large-scale evaluations both in simulation and on three different real robot embodiments. We find that as we grow and diversify its training data, RoboCat not only shows signs of cross-task transfer, but also becomes more efficient at adapting to new tasks., Comment: Transactions on Machine Learning Research (12/2023)

We build a unified framework for the study of monodromy operators and weight filtrations of cohomology theories for varieties over a local field. As an application, we give a streamlined definition of Hyodo-Kato cohomology without recourse to log-geometry, as predicted by Fontaine, and we produce an induced Clemens-Schmid chain complex., Comment: 51 pages. Minor changes to Section 4.7. To appear in JEMS

We investigate critical equilibrium and out of equilibrium properties of a ferromagnetic Ising model in one and two dimension in the presence of long range interactions, $J_{ij}\propto r^{-(d+\sigma)}$. We implement a novel local dynamics on a dynamical L\'evy lattice, that correctly reproduces the static critical exponents known in the literature, as a function of the interaction parameter $\sigma$. Due to its locality the algorithm can be applied to investigate dynamical properties, of both discrete and continuous long range models. We consider the relaxation time at the critical temperature and we measure the dynamical exponent $z$ as a function of the decay parameter $\sigma$, highlighting that the onset of short range regime for the dynamical critical properties appears to occur at a value of $\sigma$ which differs from the equilibrium one.

The ability to effectively reuse prior knowledge is a key requirement when building general and flexible Reinforcement Learning (RL) agents. Skill reuse is one of the most common approaches, but current methods have considerable limitations.For example, fine-tuning an existing policy frequently fails, as the policy can degrade rapidly early in training. In a similar vein, distillation of expert behavior can lead to poor results when given sub-optimal experts. We compare several common approaches for skill transfer on multiple domains including changes in task and system dynamics. We identify how existing methods can fail and introduce an alternative approach to mitigate these problems. Our approach learns to sequence existing temporally-extended skills for exploration but learns the final policy directly from the raw experience. This conceptual split enables rapid adaptation and thus efficient data collection but without constraining the final solution.It significantly outperforms many classical methods across a suite of evaluation tasks and we use a broad set of ablations to highlight the importance of differentc omponents of our method.

Wireless networks are commonly used in public spaces, universities and public institutions and provide accurate and easily accessible information to monitor the mobility and behavior of users. Following the application of containment measures during the recent pandemic, we analyse extensive data from the WiFi network in a University Campus in Italy during three periods, corresponding to partial lockdown, partial opening, and almost complete opening. We measure the probability distributions of groups and link activation at Wi-Fi Access Points, investigating how different areas are used in the presence of restrictions. We rank the hotspots and the area they cover according to their crowding and to the probability of link formation, which is the relevant variable in determining potential outbreaks. We consider a recently proposed epidemic model on simplicial temporal networks and we use the measured distributions to infer the change in the reproduction number in the three phases. Our data show that additional measures are necessary to limit the epidemic spreading in the total opening phase, due to the dramatic increase in the number of contacts.

We study standard and higher-order birth-death processes on fully connected networks, within the perspective of large-deviation theory (also referred to as Wentzel-Kramers-Brillouin (WKB) method in some contexts). We obtain a general expression for the leading and next-to-leading terms of the stationary probability distribution of the fraction of "active" sites, as a function of parameters and network size $N$. We reproduce several results from the literature and, in particular, we derive all the moments of the stationary distribution for the $q$-susceptible-infected-susceptible ($q-SIS$) model, i.e., a high-order epidemic model requiring of $q$ active ("infected") sites to activate an additional one. We uncover a very rich scenario for the fluctuations of the fraction of active sites, with non-trivial finite-size-scaling properties. In particular, we show that the variance-to-mean ratio diverges at criticality for $[1 \leq q\leq 3]$, with a maximal variability at $q=2$, confirming that complex-contagion processes can exhibit peculiar scaling features including wild variability and that the leading-order in a large-deviation approach does not suffice to describe them: next-to-leading terms are essential to capture the intrinsic singularity at the origin of systems with absorbing states., Comment: 10 pages, 4 figures

Knowing the exact 3D location of workers and robots in a collaborative environment enables several real applications, such as the detection of unsafe situations or the study of mutual interactions for statistical and social purposes. In this paper, we propose a non-invasive and light-invariant framework based on depth devices and deep neural networks to estimate the 3D pose of robots from an external camera. The method can be applied to any robot without requiring hardware access to the internal states. We introduce a novel representation of the predicted pose, namely Semi-Perspective Decoupled Heatmaps (SPDH), to accurately compute 3D joint locations in world coordinates adapting efficient deep networks designed for the 2D Human Pose Estimation. The proposed approach, which takes as input a depth representation based on XYZ coordinates, can be trained on synthetic depth data and applied to real-world settings without the need for domain adaptation techniques. To this end, we present the SimBa dataset, based on both synthetic and real depth images, and use it for the experimental evaluation. Results show that the proposed approach, made of a specific depth map representation and the SPDH, overcomes the current state of the art., Comment: IROS2022 and IEEE Robotics and Automation Letters (RA-L). Accepted June, 2022

We give a proof of the $p$-adic weight monodromy conjecture for scheme-theoretic complete intersections in projective smooth toric varieties. The strategy is based on Scholze's proof in the $\ell$-adic setting, which we adapt using homotopical results developed in the context of rigid analytic motives., Comment: 35 pages

In this paper, we want to speculate about the possibility to model all the currently known/proposed approaches to terminology into a single schema. We will use the Entity-Relationship (ER) diagram as our tool for the conceptual data model of the problem and to express the associations between the objects of the study. We will analyse the onomasiological and semasiological approaches, the ontoterminology paradigm, and the frame-based model, and we will draw the consequences in terms of the conceptual data model. The result of this discussion will be used as the basis of the next step of the data organization in terms of standardized terminological records and Linked Data.

For robots operating in the real world, it is desirable to learn reusable behaviours that can effectively be transferred and adapted to numerous tasks and scenarios. We propose an approach to learn abstract motor skills from data using a hierarchical mixture latent variable model. In contrast to existing work, our method exploits a three-level hierarchy of both discrete and continuous latent variables, to capture a set of high-level behaviours while allowing for variance in how they are executed. We demonstrate in manipulation domains that the method can effectively cluster offline data into distinct, executable behaviours, while retaining the flexibility of a continuous latent variable model. The resulting skills can be transferred and fine-tuned on new tasks, unseen objects, and from state to vision-based policies, yielding better sample efficiency and asymptotic performance compared to existing skill- and imitation-based methods. We further analyse how and when the skills are most beneficial: they encourage directed exploration to cover large regions of the state space relevant to the task, making them most effective in challenging sparse-reward settings.

Recently, learning frameworks have shown the capability of inferring the accurate shape, pose, and texture of an object from a single RGB image. However, current methods are trained on image collections of a single category in order to exploit specific priors, and they often make use of category-specific 3D templates. In this paper, we present an alternative approach that infers the textured mesh of objects combining a series of deformable 3D models and a set of instance-specific deformation, pose, and texture. Differently from previous works, our method is trained with images of multiple object categories using only foreground masks and rough camera poses as supervision. Without specific 3D templates, the framework learns category-level models which are deformed to recover the 3D shape of the depicted object. The instance-specific deformations are predicted independently for each vertex of the learned 3D mesh, enabling the dynamic subdivision of the mesh during the training process. Experiments show that the proposed framework can distinguish between different object categories and learn category-specific shape priors in an unsupervised manner. Predicted shapes are smooth and can leverage from multiple steps of subdivision during the training process, obtaining comparable or state-of-the-art results on two public datasets. Models and code are publicly released., Comment: Accepted at 3DV 2021