arXiv subjects

Ana Estrada-Real

Publications and source records attributed to Ana Estrada-Real.

6 recordsLinked to original sources

arXiv2026

Vision Language Model Fusion for Explainable Face Recognition

Responsible deployment of face verification systems requires more than accurate decisions: systems should also provide interpretable and auditable evidence that enables users to understand, assess, and challenge their decisions. Vision-language models (VLMs) provide a promising foundation for explainable face recognition by combining visual analysis with natural-language reasoning. However, relying on a single model may further limit the decision accuracy as well as provided explanations. This work therefore investigates whether multiple VLMs can be combined to improve recognition accuracy, and to enrich the explanations associated with those decisions. This work evaluates four VLMs as standalone face verification systems and subsequently proposes a fusion framework, where two source models provide similarity scores and textual justifications and a third VLM acts as a decider model. Four different fusion scenarios are considered, progressively providing the decider model with scores, justifications, face images, and combinations of these modalities. Overall, the findings suggest that the value of multi-VLM fusion extends beyond recognition performance. VLMs can provide complementary justifications and perspectives that enable richer explanations of face recognition decisions, supporting greater transparency, auditability, and error analysis. This is relevant to the development of responsible explainable face verification systems, where users and operators should be able to understand not only the final decision but also the evidence and potential sources underlying it. The proposed multimodal VLM, which combines decision scores, explanations, and face images, achieves higher recognition accuracy than state-of-the-art VLMs and domain-specific face recognition models, while also providing fused explanations that are expected to be more robust than those generated by individual VLMs.

cs.CV↗

arXiv2022

Probing the optical near-field interaction of Mie nanoresonators with atomically thin semiconductors

Optical Mie resonators based on silicon nanostructures allow tuning of light-matter-interaction with advanced design concepts based on CMOS compatible nanofabrication. Optically active materials such as transition-metal dichalcogenide (TMD) monolayers can be placed in the near-field region of such Mie resonators. Here, we experimentally demonstrate and verify by numerical simulations coupling between a MoSe2 monolayer and the near-field of dielectric nanoresonators. Through a comparison of dark-field (DF) scattering spectroscopy and photoluminescence excitation experiments (PLE), we show that the MoSe2 absorption can be enhanced via the near-field of a nanoresonator. We demonstrate spectral tuning of the absorption via the geometry of individual Mie resonators. We show that we indeed access the optical near-field of the nanoresonators, by measuring a spectral shift between the typical near-field resonances in PLE compared to the far-field resonances in DF scattering. Our results prove that using MoSe2 as an active probe allows accessing the optical near-field above photonic nanostructures, without the requirement of highly complex near-field microscopy equipment.

physics.optics↗

arXiv2022

Inverse design with flexible design targets via deep learning: Tailoring of electric and magnetic multipole scattering from nano-spheres

Deep learning is a promising, ultra-fast approach for inverse design in nano-optics, but despite fast advancement of the field, the computational cost of dataset generation, as well as of the training procedure itself remains a major bottleneck. This is particularly inconvenient because new data need to be generated and a new network needs to be trained for any modification of the problem. We propose a technique that allows to train a single neural network on a broad range of design targets without any re-training. The key idea of our method is to enrich existing data with random "regions of interest" (ROI) labels. A model trained on such ROI-decorated data becomes capable to operate on a broad range of physical targets, while it learns to focus its design effort on a user-defined ROI, ignoring the rest of the physical domain. We demonstrate the method by training a tandem-network on the design of dielectric core-shell nano-spheres for electric and magnetic dipole and quadrupole scattering over a broad spectral range. The network learns to tailor very distinct, flexible design targets like scattering due to specific multipoles in narrow spectral windows. Varying the design problem does not require any re-training. Our approach is very general and can be directly used with existing datasets. It can be straightforwardly applied to other network architectures and problems.

physics.optics↗

arXiv2022

One pot chemical vapor deposition of high optical quality large area monolayer Janus transition metal dichalcogenides

We report one-pot chemical vapor deposition (CVD) growth of large-area Janus SeMoS monolayers, with the asymmetric top (Se) and bottom (S) chalcogen atomic planes with respect to the central transition metal (Mo) atoms. The formation of these two-dimensional semiconductor monolayers takes place upon the thermodynamic equilibrium-driven exchange of the bottom Se atoms of the initially grown MoSe2 single crystals on gold foils with S atoms. The growth process is characterized by complementary experimental techniques including Raman and X-ray photoelectron spectroscopy and the growth mechanisms are rationalized by first principle calculations. The remarkably high optical quality of the synthesized Janus monolayers is demonstrated by optical and magneto-optical measurements which reveal the strong exciton-phonon coupling and enable to obtain the exciton g-factor of -3.3.

cond-mat.mtrl-sci↗

arXiv2022

Generalizing the exact multipole expansion: Density of multipole modes in complex photonic nanostructures

The multipole expansion of a nano-photonic structure's electromagnetic response is a versatile tool to interpret optical effects in nano-optics, but it only gives access to the modes that are excited by a specific illumination. In particular the study of various illuminations requires multiple, costly numerical simulations. Here we present a formalism we call "generalized polarizabilities", in which we combine the recently developed exact multipole decomposition [Alaee et al., Opt. Comms. 407, 17-21 (2018)] with the concept of a generalized field propagator. After an initial computation step, our approach allows to instantaneously obtain the exact multipole decomposition for any illumination. Most importantly, since all possible illuminations are included in the generalized polarizabilities, our formalism allows to calculate the total density of multipole modes, regardless of a specific illumination, which is not possible with the conventional multipole expansion. Finally, our approach directly provides the optimum illumination field distributions that maximally couple to specific multipole modes. The formalism will be very useful for various applications in nano-optics like illumination-field engineering, or meta-atom design e.g. for Huygens metasurfaces. We provide a numerical open source implementation compatible with the pyGDM python package.

physics.optics↗

arXiv2022

Exciton spectroscopy and diffusion in MoSe2-WSe2 lateral heterostructures encapsulated in hexagonal boron nitride

Chemical vapor deposition (CVD) allows lateral edge epitaxy of transition metal dichalcogenide heterostructures with potential applications in optoelectronics. Critical for carrier and exciton transport is the quality of the two materials that constitute the monolayer and the nature of the lateral heterojunction. Important details of the optical properties were inaccessible in as-grown heterostructure samples due to large inhomogeneous broadening of the optical transitions. Here we perform optical spectroscopy at T = 4 K and also at 300 K to access the optical transitions in CVD grown MoSe2-WSe2 lateral heterostructures that are transferred from the growth-substrate and are encapsulated in hBN. Photoluminescence (PL), reflectance contrast and Raman spectroscopy reveal considerably narrowed optical transition linewidth similar to high quality exfoliated monolayers. In high-resolution transmission electron microscopy (HRTEM) we find near-atomically sharp junctions with a typical extent of 3nm for the covalently bonded MoSe2-WSe2. In PL imaging experiments we find effective excitonic diffusion length that are longer for WSe2 than for MoSe2 at low T=4 K, whereas at 300 K this trend is reversed.

cond-mat.mtrl-sci↗