HAL Portal IOGS (nstitut d'Optique Graduate School)
Not a member yet
12589 research outputs found
Sort by
Les erreurs de reconstruction : une explication simple et efficace pour les auto-encodeurs de graphe utilisés pour la détection d'anomalies
No full textInternational audienceLes auto-encodeurs de graphes (Graph Auto-Encoders ou GAEs) ont fait preuve d'une efficacité remarquable pour la détection d'anomalies dans des graphes. Cependant, leur nature de "boîte noire" ne permet pas de comprendre les raisons qui les ont conduits à classer un noeud comme anormal. De plus, alors qu'avec le développement de l'XAI, de nombreuses méthodes ont été proposées pour fournir des explications pour différents modèles d'apprentissage profond, il y a une absence notable de cadre d'évaluation dédié à la détection d'anomalies dans les graphes. Notre contribution comble cette lacune en adaptant des cadres d'évaluation existants aux défis spécifiques de la détection d'anomalies à l'aide de GAEs. De plus, elle introduit une technique d'explication simple mais efficace basée sur les erreurs de reconstruction des GAEs. En utilisant ce nouveau cadre, nous évaluons l'efficacité de différents explainers et montrons expérimentalement que la méthode que nous proposons, basée sur les erreurs de reconstruction, surpasse les autres explainers pour les GAEs
Rapport de Conjoncture 2021-2025 : Section 10 : Milieux fluides et réactifs : transports, transferts, procédés de transformation
No full textComparison of experimental vs simulated data to train neural networks for speckle imaging data analysis
No full textInternational audienceMultiple Laser Speckle contrast Imaging (MESI) is an imaging method that provides relative blood flow maps from the statistical analysis of the dynamic speckle patterns observed when a coherent source is used to illuminate a tissue that contains moving scatterers. The gold standard analysis of MESI data is done by pixelwise regression of the experimental images to a theoretical function of the contrast K as a function of the exposure time T and decorrelation time tc. This approach is computer intensive, and the duration required to obtain a single flow map is too long for "real-time" analysis of in vivo hemodynamics. In addition, the mathematical model used relies on assumptions that oversimplify the local flow within the object of study. We have evaluated as an alternative a method based on Convolutional Neural Networks (CNN) to directly infer blood flow maps from MESI data, bypassing the model based fitting procedure. The CNN approach is model-free and delivers blood flow maps several orders of magnitude faster than the classical pixelwise non-linear regression. Here, we have evaluated two different datasets of annotated speckle contrast images to train the neural networks. One is composed of simulated time integrated speckle while the other one is composed of experimental data acquired for microfluidic channels with controlled geometries and flows. The study aims at discussing the assets and limits of both approaches.</div
Tailoring the visual appearance with disordered metasurfaces
No full textInternational audienceWe propose a new application of metasurfaces, namely the harnessing of the visual appearance. We demonstrate that disordered metasurfaces composed of high-index resonant metaatoms, once deposited on macroscopic objects, offer visual appearances with novel visual properties
Strong Laser Emission Modulation by Coherent Perfect Absorption inside complex-coupled Distributed Feedback Laser Diodes
No full textInternational audienceThe proof-of-concept of the exploitation of Coherent Perfect Absorption (CPA) in electrically-injected distributed-feedback laser sources is reported. Capitalizing on the essence of CPA as "light extinction by light", an integrated laser-modulator scheme emerges. The key ingredient compared to conventional single-frequency laser diodes is a careful periodic inphase modulation of both real and imaginary parts of the complex grating index profile that enables both single frequency operation and 40 dB line purity at the Bragg frequency. We show that this combination is most apt for the operation of CPA as a modulation mechanism that respects the laser spectral purity. The specific proof-of-concept is based on an ultra-short external cavity formed by a metallic micro-mirror, whose role is to generate the second beam of more conventional CPA interferometric approaches. The implemented complex-coupled grating is compatible with existing industrial technologies and promising for real-life laser source applications. Furthermore, the concept can be directly transferred to other material platforms and other wavelengths ranging from terahertz to ultraviolet
Atom interferometry in an Einstein Elevator
No full textRecent advances in atom interferometry have led to the development of quantum inertial sensors with outstanding performance in terms of sensitivity, accuracy, and long-term stability. For ground-based implementations, these sensors are ultimately limited by the free-fall height of atomic fountains required to interrogate the atoms over extended timescales. This limitation can be overcome in Space and in unique ``microgravity'' facilities such as drop towers or free-falling aircraft. These facilities require large investments, long development times, and place stringent constraints on instruments that further limit their widespread use. The available ``up time'' for experiments is also quite low, making extended studies challenging. In this work, we present a new approach in which atom interferometry is performed in a laboratory-scale Einstein Elevator. Our experiment is mounted to a moving platform that mimics the vertical free-fall trajectory every 13.5 seconds. With a total interrogation time of ms, we demonstrate an acceleration sensitivity of m/s per shot, limited primarily by the temperature of our atomic samples. We further demonstrate the capability to perform long-term statistical studies by operating the Einstein Elevator over several days with high reproducibility. These represent state-of-the-art results achieved in microgravity and further demonstrates the potential of quantum inertial sensors in Space. Our microgravity platform is both an alternative to large atomic fountains and a versatile facility to prepare future Space missions
Near-field heat transfer in many-body systems: from theory to applications
No full textInternational audienc
Maximizing the efficiency of intrapulse difference frequency generation by pulse shaping and recycling
No full textInternational audienceIntrapulse Difference Frequency Generation (iDFG) is an interesting technique for generating femtosecond pulses in the Mid-Infrared (MIR) range with unique properties such as robust Carrier-Envelope Phase (CEP) stability. However, its efficiency is low compared to other techniques. In this paper, we describe an iDFG system operating within the 4 to 10 µm range that features an original architecture to enhance efficiency. First, we introduce an interesting technique on the generation process. This approach involves polarization and spectral phase shaping techniques on the driving pulse to maximize the number of photons enrolled in the process. Second, we demonstrate that the polarization shaping allows further enhancement of efficiency by recycling the iDFG signal to pump a subsequent optical parametric amplification (OPA) stage. These two concepts and the associated parameters optimization are described into details, and supported by experimental results. Combined with a high-power Yb-fiber-based pump laser, these techniques allow to achieve record efficiencies, and generate µJ-level, fewcycle, tunable, CEP-stable pulses in the MIR at repetition rates above 100 kHz.</div
Historical Printed Ornaments: Dataset and Tasks
No full textInternational audienceThis paper aims to develop the study of historical printed ornaments with modern unsupervised computer vision. We highlight three complex tasks that are of critical interest to book historians: clustering, element discovery, and unsupervised change localization. For each of these tasks, we introduce an evaluation benchmark, and we adapt and evaluate state-of-the-art models. Our Rey's Ornaments dataset is designed to be a representative example of a set of ornaments historians would be interested in. It focuses on an XVIIIth century bookseller, Marc-Michel Rey, providing a consistent set of ornaments with a wide diversity and representative challenges. Our results highlight the limitations of state-of-the-art models when faced with real data and show simple baselines such as k-means or congealing can outperform more sophisticated approaches on such data. Our dataset and code can be found at https://printed-ornaments.github.io/
REXPACO ASDI: Joint unmixing and deconvolution of the circumstellar environment by angular and spectral differential imaging
No full textAccepté (03 Octobre 2024), Monthly Notices of the Royal Astronomical SocietyInternational audienceAngular and spectral differential imaging is an observational technique of choice to investigate the immediate vicinity of stars. By leveraging the relative angular motion and spectral scaling between on-axis and off-axis sources, post-processing techniques can separate residual star light from light emitted by surrounding objects such as circumstellar disks or point-like objects. This paper introduces a new algorithm that jointly unmixes these components and deconvolves disk images. The proposed algorithm is based on a statistical model of the residual star light, accounting for its spatial and spectral correlations. These correlations are crucial yet remain inadequately modeled by existing reconstruction algorithms. We employ dedicated shrinkage techniques to estimate the large number of parameters of our correlation model in a data-driven fashion. We show that the resulting separable model of the spatial and spectral covariances captures very accurately the star light, enabling its efficient suppression. We apply our method to datasets from the VLT/SPHERE instrument and compare its performance with standard algorithms (median subtraction, PCA, PACO). We demonstrate that considering the multiple correlations within the data significantly improves reconstruction quality, resulting in better preservation of both disk morphology and photometry. With its unique joint spectral modeling, the proposed algorithm can reconstruct disks with circular symmetry (e.g., rings, spirals) at intensities one million times fainter than the star, without needing additional reference datasets free from off-axis objects