Publications

2022

• Computational models of disfluencies : fillers and discourse markers in spoken language understanding
  
  • Dinkar Tanvi
  , 2022. People rarely speak in the same manner that they write – they are generally disfluent. Disfluencies can be defined as interruptions in the regular flow of speech, such as pausing silently, repeating words, or interrupting oneself to correct something said previously. Despite being a natural characteristic of spontaneous speech, and the rich linguistic literature that discusses their informativeness, they are often removed as noise in post-processing from the output transcripts of speech recognisers. So far, their consideration in a Spoken Language Understanding (SLU) context has been rarely explored. The aim of this thesis is to develop computational models of disfluencies in SLU. To do so, we take inspiration from psycholinguistic models of disfluencies, which focus on the role that disfluencies play in the production (by the speaker) and comprehension (by the listener) of speech. Specifically, when we use the term ``computational models of disfluencies'', we mean to develop methodologies that automatically process disfluencies to empirically observe 1) their impact on the production and comprehension of speech, and 2) how they interact with the primary signal (the lexical, or what was said in essence). To do so, we focus on two discourse contexts; monologues and task-oriented dialogues.Our results contribute to broader tasks in SLU, and also research relevant to Spoken Dialogue Systems. When studying monologues, we use a combination of traditional and neural models to study the representations and impact of disfluencies on SLU performance. Additionally, we develop methodologies to study disfluencies as a cue for incoming information in the flow of the discourse. In studying task-oriented dialogues, we focus on developing computational models to study the roles of disfluencies in the listener-speaker dynamic. We specifically study disfluencies in the context of verbal alignment; i.e. the alignment of the interlocutors' lexical expressions, and the role of disfluencies in behavioural alignment; a new alignment context that we propose to mean when instructions given by one interlocutor are followed with an action by another interlocutor. We also consider how these disfluencies in local alignment contexts can be associated with discourse level phenomena; such as success in the task. We consider this thesis one of the many first steps that could be undertaken to integrate disfluencies in SLU contexts.
• Metric learning for video to music recommendation
  
  • Prétet Laure
  , 2022. Music enhances moving images and allows to efficiently communicate emotion or narrative tension, thanks to cultural codes common to the filmmakers and viewers. A successful communication requires not only a choice of track matching the video's mood and content, but also a temporal synchronization of the audio and visual main events. This is the goal of the music supervision industry, which traditionally carries out the task manually. In this dissertation, we study the automation of tasks related to music supervision. The music supervision problem generally doesn't have a unique solution, as it includes external constraints such as the client's identity or budget. It is thus relevant to proceed by recommendation. As the number of available musical videos is in constant augmentation, it makes sense to use data-driven tools. More precisely, we use the metric learning paradigm to learn the relevant projections of multimodal (video and music) data. First, we address the music similarity problem, which is used to broaden the results of a music search. We implement an efficient content-based imitation of a tag-based similarity metric. To do so, we present a method to train a convolutional neural network from ranked lists. Then, we focus on direct, content-based music recommendation for video. We adapt a simple self-supervised system and we demonstrate a way to improve its performance, by using pretrained audio features and learning their aggregation. We then carry a qualitative and quantitative analysis of official music videos to better understand the temporal organization of musical videos. Results show that official music videos are carefully edited in order to align audio and video events, and that the level of synchronization depends on the music and video genres. With this insight, we propose the first recommendation system designed specifically for music supervision: the Seg-VM-Net, which uses both content and structure to perform the matching of music and video.
• High-definition video broadcasting with a room-temperature quantum cascade laser emitting in the long-wave infrared domain
  
  • Didier Pierre
  • Yang Ke
  • Spitz Olivier
  • Guillaume-Manca Alice
  • Liu Junqi
  • Grillot Frédéric
  , 2022, pp.46. Quantum cascade lasers (QCLs) are relevant optical sources for free-space communication because they can emit in the long-wave infrared (LWIR) domain, i.e. in the 8-12 µm region. The advantage of this optical domain is that it combines a high atmosphere transmission1 with a reduced distortion for propagating beams,2 thus the superiority of LWIR lasers in comparison with existing near-infrared systems is very dependent on link availability.3 Furthermore, QCLs are characterized by the absence of relaxation oscillation resonance.4 This peculiarity could imply a very large modulation bandwidth, even if QCL structures still need to be optimized to avoid parasitic effects.5 Recent experimental efforts have highlighted the potential of QCL-based free-space communication systems6–8 and the current 4 Gbits/s record rate is expected to be outperformed in the near future with bandwidth-enhanced structures.9 This work describes a free-space live video broadcasting with a room-temperature QCL emitting at 8.1 µm. The video file is encoded in uncompressed high-definition format (1280 pixels x 720 pixels) and this corresponds to a data rate of 1.485 Gbits/s with on-off keying scheme. This high-speed electrical signal is directly injected in the QCL via the AC port of a bias tee. The modulated optical signal from the QCL is retrieved with a Mercury-Cadmium-Telluride detector and the resulting electrical signal is sent to a TV monitor where the video can be watched in live. The current findings demonstrate the versatility of a communication system with QCLs and this paves the way for real-field applications (10.1117/12.2608511)
  DOI : 10.1117/12.2608511
• Chaos-based mid-infrared communications
  
  • Grillot Frédéric
  • Spitz Olivier
  , 2022, 11995, pp.1. The advantage of mid-infrared wavelength is that it is less affected by atmospheric conditions than conventional near-infrared wavelength, and this optical domain is thus envisioned to play a key role in the 6G standard under development. The directivity of the beam, as well as the stealth conferred by the background emission, makes communication systems based on long-wave infrared quantum cascade lasers (QCL) highly desirable. However, some applications require a further level of privacy. Protecting the communication link against eavesdroppers is possible with chaos-based enciphering. Using this concept, a chaotic master QCL is used to conceal the private message while deciphering is achieved with a second, identical, remote QCL that is called the slave. The deciphering process relies on chaos anti-synchronization where the slave only reproduces the reversed chaotic pattern of the master, thus allowing the recovery of the private message by adding the slave signal and the master signal. The privacy of our system is also assessed and shows that an illegitimate receiver would end with a detrimental error rate during translation, even in the unlikely case this eavesdropper knows the coding format of the private message. We believe our private communication system brings a cost-effective, reliable and versatile alternative for free-space data links, especially in harsh environments where mid-infrared lasers strongly outperform their near-infrared counterparts. Features such as room-temperature operation and highspeed transmission further advocates for a large deployment, and we anticipate that this finding can have a significant impact on the development of novel applications based on QCLs. (10.1117/12.2613797)
  DOI : 10.1117/12.2613797
• Effects of external optical feedback in InAs/InP quantum dot frequency comb lasers on silicon
  
  • Renaud Thibaut
  • Huang Heming
  • Liang Di
  • Kurczveil Geza
  • Beausoleil Raymond
  • Grillot Frédéric
  , 2022, 11995, pp.17. On-chip integration of semiconductor lasers have shown a growing interest in recent years, especially for the development of photonic integrated circuits (PICs) which are of paramount importance for high-speed communication within and between data centers, and fast on-board data exchanges. For all these applications, a key challenge remains the stability of the laser sources integrated on a PIC in presence of external optical feedback with the view to avoid integrated bulky and costly optical isolation. In this study, the effects of external optical feedback are investigated in hybrid InAs/InP quantum dot comb lasers on silicon. The design of the cavity includes a semiconductor optical amplifier section, a saturable absorber and an on-chip external cavity incorporating a vertical coupler. We measured the resulting feedback properties with respect to the operation conditions (bias current and voltage) and to the length of the saturable absorber. We show that under most operating conditions, the laser remains stable against optical feedback, only few regimes of operation occur, which either improve or degrade the frequency comb and/or the radio-frequency beatnote power of the laser. (10.1117/12.2608949)
  DOI : 10.1117/12.2608949
• Design territorial, représentations spatiales et participation citoyenne : revue de cas et analyse d’outils
  
  • Jolivet-Duval Marion
  • Safin Stéphane
  • Huron Samuel
  Sciences du Design, Presses Universitaires de France, 2022, n° 14 (2), pp.55-75. (10.3917/sdd.014.0055)
  DOI : 10.3917/sdd.014.0055
• Hair Color Digitization through Imaging and Deep Inverse Graphics
  
  • Kips Robin
  • Bokaris Panagiotis-Alexandros
  • Perrot Matthieu
  • Gori Pietro
  • Bloch Isabelle
  , 2022. Hair appearance is a complex phenomenon due to hair geometry and how the light bounces on different hair fibers. For this reason, reproducing a specific hair color in a rendering environment is a challenging task that requires manual work and expert knowledge in computer graphics to tune the result visually. While current hair capture methods focus on hair shape estimation many applications could benefit from an automated method for capturing the appearance of a physical hair sample, from augmented/virtual reality to hair dying development. Building on recent advances in inverse graphics and material capture using deep neural networks, we introduce a novel method for hair color digitization. Our proposed pipeline allows capturing the color appearance of a physical hair sample and renders synthetic images of hair with a similar appearance, simulating different hair styles and/or lighting environments. Since rendering realistic hair images requires path-tracing rendering, the conventional inverse graphics approach based on differentiable rendering is untractable. Our method is based on the combination of a controlled imaging device, a path-tracing renderer, and an inverse graphics model based on self-supervised machine learning, which does not require to use differentiable rendering to be trained. We illustrate the performance of our hair digitization method on both real and synthetic images and show that our approach can accurately capture and render hair color.
• Radio frequency exposure analysis in 5G massive MIMO systems
  
  • Al Hajj Maarouf
  , 2022. This dissertation presents in-situ measurements of a massive MIMO antenna and analyses the different parameters pertinent to the estimation of the EMF exposure in a 5G network. Multiple methods are presented and discussed to estimate the power received in the network while focusing on the advantages and inconveniences in each of them. This dissertation also proposes a new analytical method for studying the average exposure, presented by the total power received, in a 5G mmWave massive MIMO network. Using stochastic geometry, a close-form equation of the exposure is developed and studied by fitting a mmWave channel model using NYUSIM into statistical distributions and by modeling the BSs as a PPP. A sensitivity analysis is performed to quantify the influence of the input variables onto the exposure.Another model for a multi-user massive MIMO network is also developed deploying maximum-ratio combining precoding and max-min fairness downlink power control, and where MTs are distributed following a PPP and can be either LoS or NLoS. A closed-form expression of the expectation of the total power received and the expression of the ratio between the total power and the SIR at the nearest MT to its serving BS, where the exposure is highest. The average exposure is then studied in relation to network parameters taking into account the trade-offs presented by the power control model and antenna gains. Likewise, the ratio between the exposure and SIR is also analyzed to study the increase of exposure per the increase of the SIR at the nearest MT to its BS. And it is shown that the higher the number of antenna elements a massive MIMO antenna has, the more efficient it is in terms of SIR considering the produced exposure.
• Use of multi-temporal and multi-sensor data for continental water body extraction in the context of the SWOT mission
  
  • Gasnier Nicolas
  , 2022. Spaceborne remote sensing provides hydrologists and decision-makers with data that are essential for understanding the water cycle and managing the associated resources and risks. The SWOT satellite, which is a collaboration between the French (CNES) and American (NASA, JPL) space agencies, is scheduled for launch in 2022 and will measure the height of lakes, rivers, and oceans with high spatial resolution. It will complement existing sensors, such as the SAR and optical constellations Sentinel-1 and 2, and in situ measurements. SWOT represents a technological breakthrough as it is the first satellite to carry a near-nadir swath altimeter. The estimation of water levels is done by interferometry on the SAR images acquired by SWOT. Detecting water in these images is therefore an essential step in processing SWOT data, but it can be very difficult, especially with low signal-to-noise ratios, or in the presence of unusual radiometries. In this thesis, we seek to develop new methods to make water detection more robust. To this end, we focus on the use of exogenous data to guide detection, the combination of multi-temporal and multi-sensor data and denoising approaches. The first proposed method exploits information from the river database used by SWOT (derived from GRWL) to detect narrow rivers in the image in a way that is robust to both noise in the image, potential errors in the database, and temporal changes. This method relies on a new linear structure detector, a least-cost path algorithm, and a new Conditional Random Field segmentation method that combines data attachment and regularization terms adapted to the problem. We also proposed a method derived from GrabCut that uses an a priori polygon containing a lake to detect it on a SAR image or a time series of SAR images. Within this framework, we also studied the use of a multi-temporal and multi-sensor combination between Sentinel-1 SAR and Sentinel-2 optical images. Finally, as part of a preliminary study on denoising methods applied to water detection, we studied the statistical properties of the geometric temporal mean and proposed an adaptation of the variational method MuLoG to denoise it.
• Towards Internet Of Secure Elements
  
  • Urien Pascal
  , 2022, pp.949-950. This poster introduces the Internet Of Secure Elements (IOSE) infrastructure. IOSE enables the deployment of personal Hardware Secure Module (HSM), based on secure element (SE) grid, providing secure storage and tamper resistant computing resources (10.1109/CCNC49033.2022.9700663)
  DOI : 10.1109/CCNC49033.2022.9700663
• Demonstrating Internet Of Secure Elements Server
  
  • Urien Pascal
  , 2022, pp.923-924. This demonstration presents the first IOSE (Internet Of Secure Elements) server working within a Raspberry Pi, and managing four secure elements. This server executes two TCP daemons. RACS port (Remote APDU Call Secure) is used to download applications in secure elements according to global platform (GP) protocols. TLS port is used to route TLS messages to TLS-SE applications, identified by their server name. TLS-SE is an embedded software for secure element, protected by TLS1.3 secure channel (i.e. TLS server). The demo shows on-demand deployment of TLS-SE App from service provider to IOSE server. It presents the user’s enrollment procedure required to transfer the exclusive TLS-SE ownership to user. Finally it shows a use case in which TLS-SE App is used for trusted and secure Ethereum transaction generation (10.1109/CCNC49033.2022.9700553)
  DOI : 10.1109/CCNC49033.2022.9700553
• The Dichotomy of Evaluating Homomorphism-Closed Queries on Probabilistic Graphs
  
  • Amarilli Antoine
  • Ceylan İsmail İlkan
  Logical Methods in Computer Science, Logical Methods in Computer Science Association, 2022. We study the problem of query evaluation on probabilistic graphs, namely, tuple-independent probabilistic databases over signatures of arity two. We focus on the class of queries closed under homomorphisms, or, equivalently, the infinite unions of conjunctive queries. Our main result states that the probabilistic query evaluation problem is #P-hard for all unbounded queries from this class. As bounded queries from this class are equivalent to a union of conjunctive queries, they are already classified by the dichotomy of Dalvi and Suciu (2012). Hence, our result and theirs imply a complete data complexity dichotomy, between polynomial time and #P-hardness, on evaluating homomorphism-closed queries over probabilistic graphs. This dichotomy covers in particular all fragments of infinite unions of conjunctive queries over arity-two signatures, such as negation-free (disjunctive) Datalog, regular path queries, and a large class of ontology-mediated queries. The dichotomy also applies to a restricted case of probabilistic query evaluation called generalized model counting, where fact probabilities must be 0, 0.5, or 1. We show the main result by reducing from the problem of counting the valuations of positive partitioned 2-DNF formulae, or from the source-to-target reliability problem in an undirected graph, depending on properties of minimal models for the query. (10.46298/lmcs-18(1:2)2022)
  DOI : 10.46298/lmcs-18(1:2)2022
• An eager splitting strategy for online decision trees in ensembles
  
  • Manapragada Chaitanya
  • Gomes Heitor Murilo
  • Salehi Mahsa
  • Bifet Albert
  • Webb Geoffrey I.
  Data Mining and Knowledge Discovery, Springer, 2022, 36 (2), pp.566--619. Decision tree ensembles are widely used in practice. In this work, we study in ensemble settings the effectiveness of replacing the split strategy for the state-of-the-art online tree learner, Hoeffding Tree, with a rigorous but more eager splitting strategy that we had previously published as Hoeffding AnyTime Tree. Hoeffding AnyTime Tree (HATT), uses the Hoeffding Test to determine whether the current best candidate split is superior to the current split, with the possibility of revision, while Hoeffding Tree aims to determine whether the top candidate is better than the second best and if a test is selected, fixes it for all posterity. HATT converges to the ideal batch tree while Hoeffding Tree does not. We find that HATT is an efficacious base learner for online bagging and online boosting ensembles. On UCI and synthetic streams, HATT as a base learner outperforms HT at a 0.05 significance level for the majority of tested ensembles on what we believe is the largest and most comprehensive set of testbenches in the online learning literature. Our results indicate that HATT is a superior alternative to Hoeffding Tree in a large number of ensemble settings. (10.1007/S10618-021-00816-X)
  DOI : 10.1007/S10618-021-00816-X
• Approximating Perfect Recall when Model Checking Strategic Abilities: Theory and Applications
  
  • Belardinelli Francesco
  • Lomuscio Alessio
  • Malvone Vadim
  • Yu Emily
  Journal of Artificial Intelligence Research, Association for the Advancement of Artificial Intelligence, 2022, 73, pp.897-932. The model checking problem for multi-agent systems against specifications in the alternating-time temporal logic AT L, hence AT L∗ , under perfect recall and imperfect information is known to be undecidable. To tackle this problem, in this paper we investigate a notion of bounded recall under incomplete information. We present a novel three-valued semantics for AT L∗ in this setting and analyse the corresponding model checking problem. We show that the three-valued semantics here introduced is an approximation of the classic two-valued semantics, then give a sound, albeit partial, algorithm for model checking two-valued perfect recall via its approximation as three-valued bounded recall. Finally, we extend MCMAS, an open-source model checker for AT L and other agent specifications, to incorporate bounded recall; we illustrate its use and present experimental results. (10.1613/jair.1.12539)
  DOI : 10.1613/jair.1.12539
• Competition between Entrainment Phenomenon and Chaos in a Quantum-Cascade Laser under Strong Optical Reinjection
  
  • Spitz Olivier
  • Durupt Lauréline
  • Grillot Frédéric
  Photonics, MDPI, 2022, 9 (1), pp.1-10. The topic of external optical feedback in quantum-cascade lasers is relevant for stability and beam-properties considerations. Albeit less sensitive to external optical feedback than other lasers, quantum-cascade lasers can exhibit several behaviors under such feedback, and those are relevant for a large panel of applications, from communication to ranging and sensing. This work focused on a packaged Fabry–Perot quantum-cascade laser under strong external optical feedback and shows the influence of the beam-splitter characteristics on the optical power properties of this commercially available laser. The packaged quantum-cascade laser showed extended conditions of operation when subject to strong optical feedback, and the maximum power that can be extracted from the external cavity was also increased. When adding a periodic electrical perturbation, various non-linear dynamics were observed, and this complements previous efforts about the entrainment phenomenon in monomode quantum-cascade lasers, with the view of optimizing private communication based on mid-infrared quantum-cascade lasers. Overall, this work is a step forward in understanding the behavior of the complex quantum-cascade-laser structure when it is subjected to external optical feedback. (10.3390/photonics9010029)
  DOI : 10.3390/photonics9010029
• One Versus all for deep Neural Network Incertitude (OVNNI) quantification
  
  • Franchi Gianni
  • Bursuc Andrei
  • Aldea Emanuel
  • Dubuisson Séverine
  • Bloch Isabelle
  IEEE Access, IEEE, 2022. Deep neural networks (DNNs) are powerful learning models yet their results are not always reliable. This is due to the fact that modern DNNs are usually uncalibrated and we cannot characterize their epistemic uncertainty. In this work, we propose a new technique to quantify the epistemic uncertainty of data easily. This method consists in mixing the predictions of an ensemble of DNNs trained to classify One class vs All the other classes (OVA) with predictions from a standard DNN trained to perform All vs All (AVA) classification. On the one hand, the adjustment provided by the AVA DNN to the score of the base classifiers allows for a more fine-grained inter-class separation. On the other hand, the two types of classifiers enforce mutually their detection of out-of-distribution (OOD) samples, circumventing entirely the requirement of using such samples during training. Our method achieves state of the art performance in quantifying OOD data across multiple datasets and architectures while requiring little hyper-parameter tuning. (10.1109/access.2021.3138978)
  DOI : 10.1109/access.2021.3138978
• Mid-infrared hyperchaos of interband cascade lasers
  
  • Deng Yu
  • Fan Zhuo-Fei
  • Zhao Bin-Bin
  • Wang Xing-Guang
  • Zhao Shiyuan
  • Wu Jiagui
  • Grillot Frédéric
  • Wang Cheng
  Light: Science and Applications, Nature Publishing Group, 2022, 11 (7), pp.1-10. Abstract Chaos in nonlinear dynamical systems is featured with irregular appearance and with high sensitivity to initial conditions. Near-infrared light chaos based on semiconductor lasers has been extensively studied and has enabled various applications. Here, we report a fully-developed hyperchaos in the mid-infrared regime, which is produced from interband cascade lasers subject to the external optical feedback. Lyapunov spectrum analysis demonstrates that the chaos exhibits three positive Lyapunov exponents. Particularly, the chaotic signal covers a broad frequency range up to the GHz level, which is two to three orders of magnitude broader than existed mid-infrared chaos solutions. The interband cascade lasers produce either periodic oscillations or low-frequency fluctuations before bifurcating to hyperchaos. This hyperchaos source is valuable for developing long-reach secure optical communication links and remote chaotic Lidar systems, taking advantage of the high-transmission windows of the atmosphere in the mid-infrared regime. (10.1038/s41377-021-00697-1)
  DOI : 10.1038/s41377-021-00697-1
• Uniform Reliability of Self-Join-Free Conjunctive Queries
  
  • Amarilli Antoine
  • Kimelfeld Benny
  Logical Methods in Computer Science, Logical Methods in Computer Science Association, 2022. The reliability of a Boolean Conjunctive Query (CQ) over a tuple-independent probabilistic database is the probability that the CQ is satisfied when the tuples of the database are sampled one by one, independently, with their associated probability. For queries without self-joins (repeated relation symbols), the data complexity of this problem is fully characterized by a known dichotomy: reliability can be computed in polynomial time for hierarchical queries, and is #P-hard for non-hierarchical queries. Inspired by this dichotomy, we investigate a fundamental counting problem for CQs without self-joins: how many sets of facts from the input database satisfy the query? This is equivalent to the uniform case of the query reliability problem, where the probability of every tuple is required to be 1/2. Of course, for hierarchical queries, uniform reliability is solvable in polynomial time, like the reliability problem. We show that being hierarchical is also necessary for this tractability (under conventional complexity assumptions). In fact, we establish a generalization of the dichotomy that covers every restricted case of reliability in which the probabilities of tuples are determined by their relation. (10.46298/lmcs-18(4:3)2022)
  DOI : 10.46298/lmcs-18(4:3)2022
• Reasoning about Moving Target Defense in Attack Modeling Formalisms
  
  • Ballot Gabriel
  • Malvone Vadim
  • Leneutre Jean
  • Borde Etienne
  , 2022. Since 2009, Moving Target Defense (MTD) has become a new paradigm of defensive mechanism that frequently changes the state of the target system to confuse the attacker. This frequent change is costly and leads to a trade-off between misleading the attacker and disrupting the quality of service. Optimizing the MTD activation frequency is necessary to develop this defense mechanism when facing realistic, multi-step attack scenarios. Attack modeling formalisms based on DAG are prominently used to specify these scenarios. Our contribution is a new DAG-based formalism for MTDs and its translation into a Price Timed Markov Decision Process to find the best activation frequencies against the attacker's time/cost-optimal strategies. For the first time, MTD activation frequencies are analyzed in a state-of-the-art DAG-based representation. Moreover, this is the first paper that considers the specificity of MTDs in the automatic analysis of attack modeling formalisms. Finally, we present some experimental results using Uppaal Stratego to demonstrate its applicability and relevance.
• Towards Outdoor Electromagnetic Field Exposure Mapping Generation Using Conditional GANs
  
  • Mallik Mohammed
  • Tesfay Angesom Ataklity
  • Allaert Benjamin
  • Kassi Rédha
  • Egea-Lopez Esteban
  • Molina-Garcia-Pardo Jose-Maria
  • Wiart Joe
  • Gaillot Davy
  • Clavier Laurent
  Sensors, MDPI, 2022, 22 (24), pp.9643. With the ongoing fifth-generation cellular network (5G) deployment, electromagnetic field exposure has become a critical concern. However, measurements are scarce, and accurate electromagnetic field reconstruction in a geographic region remains challenging. This work proposes a conditional generative adversarial network to address this issue. The main objective is to reconstruct the electromagnetic field exposure map accurately according to the environment’s topology from a few sensors located in an outdoor urban environment. The model is trained to learn and estimate the propagation characteristics of the electromagnetic field according to the topology of a given environment. In addition, the conditional generative adversarial network-based electromagnetic field mapping is compared with simple kriging. Results show that the proposed method produces accurate estimates and is a promising solution for exposure map reconstruction. (10.3390/s22249643)
  DOI : 10.3390/s22249643
• Real-time Virtual-Try-On from a Single Example Image through Deep Inverse Graphics and Learned Differentiable Renderers
  
  • Kips Robin
  • Jiang Ruowei
  • Ba Sileye
  • Duke Brendan
  • Perrot Matthieu
  • Gori Pietro
  • Bloch Isabelle
  Computer Graphics Forum, Wiley, 2022. Augmented reality applications have rapidly spread across online platforms, allowing consumers to virtually try-on a variety of products, such as makeup, hair dying, or shoes. However, parametrizing a renderer to synthesize realistic images of a given product remains a challenging task that requires expert knowledge. While recent work has introduced neural rendering methods for virtual try-on from example images, current approaches are based on large generative models that cannot be used in real-time on mobile devices. This calls for a hybrid method that combines the advantages of computer graphics and neural rendering approaches. In this paper we propose a novel framework based on deep learning to build a real-time inverse graphics encoder that learns to map a single example image into the parameter space of a given augmented reality rendering engine. Our method leverages self-supervised learning and does not require labeled training data which makes it extendable to many virtual try-on applications. Furthermore, most augmented reality renderers are not differentiable in practice due to algorithmic choices or implementation constraints to reach real-time on portable devices. To relax the need for a graphics-based differentiable renderer in inverse graphics problems, we introduce a trainable imitator module. Our imitator is a generative network that learns to accurately reproduce the behavior of a given non-differentiable renderer. We propose a novel rendering sensitivity loss to train the imitator, which ensures that the network learns an accurate and continuous representation for each rendering parameter. Our framework enables novel applications where consumers can virtually try-on a novel unknown product from an inspirational reference image on social media. It can also be used by graphics artists to automatically create realistic rendering from a reference product image.
• A variant of Young’s method in voting theory providing the same winners as Copeland’s method
  
  • Hudry Olivier
  International Journal of Mathematics, Statistics and Operations Research, Academic Research Foundations, 2022, 1 (2), pp.101-107.
• Generalized Fast Multichannel Nonnegative Matrix Factorization Based on Gaussian Scale Mixtures for Blind Source Separation
  
  • Fontaine Mathieu
  • Sekiguchi Kouhei
  • Nugraha Aditya
  • Bando Yoshiaki
  • Yoshii Kazuyoshi
  IEEE/ACM Transactions on Audio, Speech and Language Processing, Institute of Electrical and Electronics Engineers, 2022, pp.1-1. This paper describes heavy-tailed extensions of a state-of-the-art versatile blind source separation method called fast multichannel nonnegative matrix factorization (FastMNMF) from a unified point of view. The common way of deriving such an extension is to replace the multivariate complex Gaussian distribution in the likelihood function with its heavy-tailed generalization, e.g., the multivariate complex Student's t and leptokurtic generalized Gaussian distributions, and tailor-make the corresponding parameter optimization algorithm. Using a wider class of heavy-tailed distributions called a Gaussian scale mixture (GSM), i.e., a mixture of Gaussian distributions whose variances are perturbed by positive random scalars called impulse variables, we propose GSM-FastMNMF and develop an expectationmaximization algorithm that works even when the probability density function of the impulse variables have no analytical expressions. We show that existing heavy-tailed FastMNMF extensions are instances of GSM-FastMNMF and derive a new instance based on the generalized hyperbolic distribution that include the normal-inverse Gaussian, Student's t, and Gaussian distributions as the special cases. Our experiments show that the normalinverse Gaussian FastMNMF outperforms the state-of-the-art FastMNMF extensions and ILRMA model in speech enhancement and separation in terms of the signal-to-distortion ratio. (10.1109/TASLP.2022.3172631)
  DOI : 10.1109/TASLP.2022.3172631
• La fève du boulanger
  
  • Zayana Karim
  • Michalak Pierre
  • Bréheret Richard
  • Boyer Ivan
  CultureMath, ENS, 2022.
• Some Rainbow Problems in Graphs Have Complexity Equivalent to Satisfiability Problems
  
  • Hudry Olivier
  • Lobstein Antoine
  International Transactions in Operational Research, Wiley, 2022, 29 (3), pp.1547-1572. In a vertex-coloured graph, a set of vertices S is said to be a rainbow set if every colour in the graph appears exactly once in S. We investigate the complexities of various problems dealing with domination in vertex-coloured graphs (existence of rainbow dominating sets, of rainbow locating-dominating sets, of rainbow identifying sets), including when we ask for a unique solution: we show equivalence between these complexities and those of the well-studied Boolean satisfiability problems. (10.1111/itor.12847)
  DOI : 10.1111/itor.12847