Search

In cinema, visual motifs are recurrent iconographic compositions that carry artistic or aesthetic significance. Their use throughout the history of visual arts and media is interesting to researchers and filmmakers alike. Our goal in this work is to recognise and classify these motifs by proposing a new machine learning model that uses a custom dataset to that end. We show how features extracted from a CLIP model can be leveraged by using a shallow network and an appropriate loss to classify images into 20 different motifs, with surprisingly good results: an $F_1$-score of 0.91 on our test set. We also present several ablation studies justifying the input features, architecture and hyperparameters used., Comment: 17 pages, 11 figures, one table, to be published in the conference proceedings of ECCV 2024

In this note we correct the proof of Proposition 4 in our paper "The Laplacian Spectrum of Large Graphs Sampled from Graphons" (arXiv:2004.09177) and we improve several results therein. To this end, we prove a new concentration lemma about degrees and Laplacian eigenvalues. This lemma allows us to improve several bounds and to dispense from assuming that the graphon is bounded away from zero in several results. This extension leads, in particular, to correct the proof of Proposition 4. Additionally, we extend Proposition 4 to graphs that are sampled from graphons by using deterministic latent variables., Comment: 9 pages, notation from original paper https://doi.org/10.48550/arXiv.2004.09177

We present an updated catalog of sources detected by the Mikhail Pavlinsky ART-XC telescope aboard the Spektrum-Roentgen-Gamma (SRG) observatory during its all-sky survey. It is based on the data of the first four and the partially completed fifth scans of the sky (ARTSS1-5). The catalog comprises 1545 sources detected in the 4-12 keV energy band. The achieved sensitivity ranges between $\sim 4\times 10^{-12}$ erg s$^{-1}$ cm$^{-2}$ near the ecliptic plane and $\sim 7\times 10^{-13}$ erg s$^{-1}$ cm$^{-2}$ near the ecliptic poles, which is a $\sim$30-50% improvement over the previous version of the catalog based on the first two all-sky scans (ARTSS12). There are $\sim 130$ objects, excluding the expected contribution of spurious detections, that were not known as X-ray sources before the SRG/ART-XC all-sky survey. We provide information, partly based on our ongoing follow-up optical spectroscopy program, on the identification and classification of the majority of the ARTSS1-5 sources (1463), of which 173 are tentative at the moment. The majority of the classified objects (964) are extragalactic, a small fraction (30) are located in the Local Group of galaxies, and 469 are Galactic. The dominant classes of objects in the catalog are active galactic nuclei (911) and cataclysmic variables (192)., Comment: Accepted for publication in A&A, 67 pages, 11 figures, the catalog of sources is included

In this paper we make use of graphon theory to study opinion dynamics on large undirected networks. The opinion dynamics models that we take into consideration allow for negative interactions between the individuals, i.e. competing entities whose opinions can grow apart. We consider both the repelling model and the opposing model that are studied in the literature. We define the repelling and the opposing dynamics on graphons and we show that their initial value problem's solutions exist and are unique. We then show that the graphon dynamics well approximate the dynamics on large graphs that converge to a graphon. This result applies to large random graphs that are sampled according to a graphon. All these facts are illustrated in an extended numerical example., Comment: 8 double-column pages. This revised version corrects several typos. An abridged version is going to appear in the proceedings of the 2024 IEEE Conference on Decision and Control

The cross dock door design problem consists of deciding on the number and capacity of inbound and outbound doors for receiving product pallets from origin nodes and exiting them to destination nodes. The uncertainty, realized in scenarios, lies in the occurrence of these nodes, the number and cost of the pallets, and the disruption of the capacity of the doors. It is represented using a stochastic two stage binary quadratic model. The first stage decisions are related to the cross dock infrastructure design, and the second stage decisions are related to the node to assignments of the doors. This is the first time, as far as we know, that a stochastic two stage binary quadratic model has been presented for minimizing the construction cost of the infrastructure and its exploitation expected cost in the scenarios. Given the difficulty of solving this combinatorial problem, a mathematically equivalent mixed integer linear formulation is introduced. However, searching an optimal solution is still impractical for commercial solvers. Thus, a scenario cluster decomposition based matheuristic algorithm is introduced to obtain feasible solutions with small optimality gap and reasonable computational effort. A broad study to validate the proposal gives solutions with a much smaller gap than the ones provided by a state of the art general solver. In fact, the proposal provides solutions with a 1 to 5% optimality gap, while the solver does it with up to a 12% gap, if any, and requires a wall time two orders of magnitude higher., Comment: 20 pages, 5 figures

Estimating the relative camera pose from $n \geq 5$ correspondences between two calibrated views is a fundamental task in computer vision. This process typically involves two stages: 1) estimating the essential matrix between the views, and 2) disambiguating among the four candidate relative poses that satisfy the epipolar geometry. In this paper, we demonstrate a novel approach that, for the first time, bypasses the second stage. Specifically, we show that it is possible to directly estimate the correct relative camera pose from correspondences without needing a post-processing step to enforce the cheirality constraint on the correspondences. Building on recent advances in certifiable non-minimal optimization, we frame the relative pose estimation as a Quadratically Constrained Quadratic Program (QCQP). By applying the appropriate constraints, we ensure the estimation of a camera pose that corresponds to a valid 3D geometry and that is globally optimal when certified. We validate our method through exhaustive synthetic and real-world experiments, confirming the efficacy, efficiency and accuracy of the proposed approach. Code is available at https://github.com/javrtg/C2P., Comment: Accepted to CVPR 2024

As the adoption of Artificial Intelligence (AI) systems within the clinical environment grows, limitations in bandwidth and compute can create communication bottlenecks when streaming imaging data, leading to delays in patient care and increased cost. As such, healthcare providers and AI vendors will require greater computational infrastructure, therefore dramatically increasing costs. To that end, we developed ISLE, an intelligent streaming framework for high-throughput, compute- and bandwidth- optimized, and cost effective AI inference for clinical decision making at scale. In our experiments, ISLE on average reduced data transmission by 98.02% and decoding time by 98.09%, while increasing throughput by 2,730%. We show that ISLE results in faster turnaround times, and reduced overall cost of data, transmission, and compute, without negatively impacting clinical decision making using AI systems., Comment: 5 pages, 3 figures, 3 tables

Current deep visual local feature detectors do not model the spatial uncertainty of detected features, producing suboptimal results in downstream applications. In this work, we propose two post-hoc covariance estimates that can be plugged into any pretrained deep feature detector: a simple, isotropic covariance estimate that uses the predicted score at a given pixel location, and a full covariance estimate via the local structure tensor of the learned score maps. Both methods are easy to implement and can be applied to any deep feature detector. We show that these covariances are directly related to errors in feature matching, leading to improvements in downstream tasks, including solving the perspective-n-point problem and motion-only bundle adjustment. Code is available at https://github.com/javrtg/DAC

Starting from a stochastic individual-based description of an SIS epidemic spreading on a random network, we study the dynamics when the size $n$ of the network tends to infinity. We recover in the limit an infinite-dimensional integro-differential equation studied by Delmas, Dronnier and Zitt (2022) for an SIS epidemic propagating on a graphon. Our work covers the case of dense and sparse graphs, provided that the number of edges grows faster than $n$, but not the case of very sparse graphs with $O(n)$ edges. In order to establish our limit theorem, we have to deal with both the convergence of the random graphs to the graphon and the convergence of the stochastic process spreading on top of these random structures: in particular, we propose a coupling between the process of interest and an epidemic that spreads on the complete graph but with a modified infection rate. Keywords: Random graph, mathematical models of epidemics, measure-valued process, large network limit, limit theorem, graphon., Comment: This work was financed by the Labex B\'ezout (ANR-10-LABX-58) and the COCOON grant (ANR-22-CE48-0011), and by the platform MODCOV19 of the CNRS. V.C.T. is partly financed by the Chaire "Mod\'elisation Math\'ematique et Biodiversit\'e''. The research leading to this article was largely performed while A. Velleret was a researcher at LAMA and at GIPSA-Lab

We provide a naturally isomorphic description of the persistence map from merge trees to barcodes in terms of a monotone map from the partition lattice to the subset lattice. Our description is local, which offers the potential to speed up inverse computations, and brings classical tools in combinatorics to bear on an active area of research in topological data analysis (TDA)., Comment: 4 pages, 2 figures

The International Fusion Materials Irradiation Facility (IFMIF), presently in its Engineering Validation and Engineering Design Activities (EVEDA) phase under the frame of the Broader Approach Agreement between Europe and Japan, accomplished in summer 2013, on schedule, its EDA phase with the release of the engineering design report of the IFMIF plant, which is here described. Many improvements of the design from former phases are implemented, particularly a reduction of beam losses and operational costs thanks to the superconducting accelerator concept, the re-location of the quench tank outside the test cell (TC) with a reduction of tritium inventory and a simplification on its replacement in case of failure, the separation of the irradiation modules from the shielding block gaining irradiation flexibility and enhancement of the remote handling equipment reliability and cost reduction, and the water cooling of the liner and biological shielding of the TC, enhancing the efficiency and economy of the related sub-systems. In addition, the maintenance strategy has been modified to allow a shorter yearly stop of the irradiation operations and a more careful management of the irradiated samples. The design of the IFMIF plant is intimately linked with the EVA phase carried out since the entry into force of IFMIF/EVEDA in June 2007.

We use tools from geometric group theory to produce a stratification of the space $\mathcal{B}_n$ of barcodes with $n$ bars. The top-dimensional strata are indexed by permutations associated to barcodes as defined by Kanari, Garin and Hess. More generally, the strata correspond to marked double cosets of parabolic subgroups of the symmetric group $Sym_n$. This subdivides $\mathcal{B}_n$ into regions that consist of barcodes with the same averages and standard deviations of birth and death times and the same permutation type. We obtain coordinates that form a new invariant of barcodes, extending the one of Kanari-Garin-Hess. This description also gives rise to metrics on $\mathcal{B}_n$ that coincide with modified versions of the bottleneck and Wasserstein metrics.

Digital images enable quantitative analysis of material properties at micro and macro length scales, but choosing an appropriate resolution when acquiring the image is challenging. A high resolution means longer image acquisition and larger data requirements for a given sample, but if the resolution is too low, significant information may be lost. This paper studies the impact of changes in resolution on persistent homology, a tool from topological data analysis that provides a signature of structure in an image across all length scales. Given prior information about a function, the geometry of an object, or its density distribution at a given resolution, we provide methods to select the coarsest resolution yielding results within an acceptable tolerance. We present numerical case studies for an illustrative synthetic example and samples from porous materials where the theoretical bounds are unknown., Comment: accepted for the IEEE Big Data 2021 workshop: Applications of Topological Data Analysis to 'Big Data'

Persistent Homology (PH) has been successfully used to train networks to detect curvilinear structures and to improve the topological quality of their results. However, existing methods are very global and ignore the location of topological features. In this paper, we remedy this by introducing a new filtration function that fuses two earlier approaches: thresholding-based filtration, previously used to train deep networks to segment medical images, and filtration with height functions, typically used to compare 2D and 3D shapes. We experimentally demonstrate that deep networks trained using our PH-based loss function yield reconstructions of road networks and neuronal processes that reflect ground-truth connectivity better than networks trained with existing loss functions based on PH. Code is available at https://github.com/doruk-oner/PH-TopoLoss.

We review epidemiological models for the propagation of the COVID-19 pandemic during the early months of the outbreak: from February to May 2020. The aim is to propose a methodological review that highlights the following characteristics: (i) the epidemic propagation models, (ii) the modeling of intervention strategies, (iii) the models and estimation procedures of the epidemic parameters and (iv) the characteristics of the data used. We finally selected 80 articles from open access databases based on criteria such as the theoretical background, the reproducibility, the incorporation of interventions strategies, etc. It mainly resulted to phenomenological, compartmental and individual-level models. A digital companion including an online sheet, a Kibana interface and a markdown document is proposed. Finally, this work provides an opportunity to witness how the scientific community reacted to this unique situation.

In this paper we consider two aspects of the inverse problem of how to construct merge trees realizing a given barcode. Much of our investigation exploits a recently discovered connection between the symmetric group and barcodes in general position, based on the simple observation that death order is a permutation of birth order. The first important outcome of our study is a clear combinatorial distinction between the space of phylogenetic trees (as defined by Billera, Holmes and Vogtmann) and the space of merge trees. Generic BHV trees on $n+1$ leaf nodes fall into $(2n-1)!!$ distinct strata, but the analogous number for merge trees is equal to the number of maximal chains in the lattice of partitions, i.e., $(n+1)!n!2^{-n}$. The second aspect of our study is the derivation of precise formulas for the distribution of tree realization numbers (the number of merge trees realizing a given barcode) when we assume that barcodes are sampled using a uniform distribution on the symmetric group. We are able to characterize some of the higher moments of this distribution, thanks in part to a reformulation in terms of Dirichlet convolution. This characterization provides a type of null hypothesis, apparently different from the distributions observed in real neuron data and opens the door to doing more precise science., Comment: 39 pages, 17 figures. Fixed authors' names in v2 with no other changes

In dynamic settings each economic agent's choices can be revealing of her private information. This elicitation via the rationalization of observable behavior depends each agent's perception of which payoff-relevant contingencies other agents persistently deem as impossible. We formalize the potential heterogeneity of these perceptions as disagreements at higher-orders about the set of payoff states of a dynamic game. We find that apparently negligible disagreements greatly affect how agents interpret information and assess the optimality of subsequent behavior: When knowledge of the state space is only 'almost common', strategic uncertainty may be greater when choices are rationalized than when they are not--forward and backward induction predictions, respectively, and while backward induction predictions are robust to small disagreements about the state space, forward induction predictions are not. We also prove that forward induction predictions always admit unique selections a la Weinstein and Yildiz (2007) (also for spaces not satisfying richness) and backward induction predictions do not.

ART-XC (Astronomical Roentgen Telescope - X-ray Concentrator) is the hard X-ray instrument with grazing incidence imaging optics on board the Spektr-Roentgen-Gamma (SRG) observatory. The SRG observatory is the flagship astrophysical mission of the Russian Federal Space Program, which was successively launched into orbit around the second Lagrangian point (L2) of the Earth-Sun system with a Proton rocket from the Baikonur cosmodrome on 13 July 2019. The ART-XC telescope will provide the first ever true imaging all-sky survey performed with grazing incidence optics in the 4-30 keV energy band and will obtain the deepest and sharpest map of the sky in the energy range of 4-12 keV. Observations performed during the early calibration and performance verification phase as well as during the on-going all-sky survey that started on 12 Dec. 2019 have demonstrated that the in-flight characteristics of the ART-XC telescope are very close to expectations based on the results of ground calibrations. Upon completion of its 4-year all-sky survey, ART-XC is expected to detect ~5000 sources (~3000 active galactic nuclei, including heavily obscured ones, several hundred clusters of galaxies, ~1000 cataclysmic variables and other Galactic sources), and to provide a high-quality map of the Galactic background emission in the 4-12 keV energy band. ART-XC is also well suited for discovering transient X-ray sources. In this paper, we describe the telescope, results of its ground calibrations, major aspects of the mission, the in-flight performance of ART-XC and first scientific results., Comment: 19 pages, 30 figures, accepted for publication in Astronomy and Astrophysics

To compute the persistent homology of a grayscale digital image one needs to build a simplicial or cubical complex from it. For cubical complexes, the two commonly used constructions (corresponding to direct and indirect digital adjacencies) can give different results for the same image. The two constructions are almost dual to each other, and we use this relationship to extend and modify the cubical complexes to become dual filtered cell complexes. We derive a general relationship between the persistent homology of two dual filtered cell complexes, and also establish how various modifications to a filtered complex change the persistence diagram. Applying these results to images, we derive a method to transform the persistence diagram computed using one type of cubical complex into a persistence diagram for the other construction. This means software for computing persistent homology from images can now be easily adapted to produce results for either of the two cubical complex constructions without additional low-level code implementation.