Research Publications

Permanent URI for this collection

Browse

Recent Submissions

Now showing 1 - 5 of 1137
  • Item
    Constrained Zero-Shot Neural Architecture Search on Small Classification Dataset
    (2024-05-31) Vuagniaux, Rémy; Narduzzi, Simon; Maamari, Nadim; Dunbar, L. Andrea
    The rapid evolution of DL has brought about significant transformations across scientific domains, marked by the development of increasingly intricate models demanding powerful GPU platforms. However, edge applications like wearables and monitoring systems impose stringent constraints on memory, size, and energy, making on-device processing imperative. To address these constraints, we employ an efficient zero-shot data-dependent NAS strategy, enhancing the search speed through the utilization of proxy functions. Additionally, we integrate KD during the learning process, harnessing insights from pre-trained models to enrich the performance and adaptability of our approach. This combined method not only achieves improved accuracy with but also results in a reduced memory footprint for the model. Our validation on CUB200-2011 demonstrates the feasibility of achieving a competitive NAS-optimized architecture for small datasets, compared to models pre-trained on larger ones.
  • Item
    VIVALDI: Steel Billet Tracking with Advanced Computer Vision
    (2024-03-13) Blanc, Sebastien; Pad, Pedram; Türetken, Engin; Cantale, Nicolas; Dia, Mohamad; Honzatko, David; Kündig, Clément; Dunbar, Andrea; Feldhaus, Stephan; Hauenstein, Gian; Meier, Marcel; Meier, Thomas; Mendler, Tomaso; Michelon, Giovanni
    Our advanced computer vision system allows for the precise tracking of serial numbers on steel billets in challenging industrial settings. It combines cutting-edge hardware and machine learning, excelling in character recognition (99.8%) and localization while adapting to dynamic ambient lighting conditions (10^4). Moreover, it accurately measures crucial geometric parameters such as side sizes, bulging, and skewness. This multifaceted technology promises to elevate material tracking, quality assessment, and production optimization in the steel industry to unprecedented levels.
  • Item
    Thin-film lithium niobate PICs: advancements and potential applications in telecom and beyond
    (2024-04-07) Sattari, Hamed; Despont, Michel; Prieto, Ivan; Dubois , Florian; Zarebidaki, Homa; Jacopo, Leo; Choong, G; Arefi, F; Orvietani, M; Della Torre, A; Mettraux, A; Herle, D; Petremand, Y; Palmieri, M; Dubochet, O
    The emergent thin-film lithium niobate on insulator (TFLN or LNOI) photonic integrated circuits (PICs) offer significant advantages in various applications due to their unique properties. This paper briefly explores recent advancements in TFLN PIC developments and their broad applications, emphasizing transformative capabilities in telecommunications and beyond. We highlight CSEM's pioneering initiative in establishing the first open access foundry for this technology, addressing challenges associated with limited access to manufacturing facilities and process design kits (PDK).
  • Item
    An Open-Access Platform for Thin Film Lithium Niobate Photonic Integrated Circuits (TFLN PICs)
    (2024) Sattari, Hamed; Prieto, Ivan; Zarebidaki,, Homa; Jacopo, Leo; Choong, Gregory; Arefi, Fatemeh; Orvietani, Mattia; Della Torre, Alberto; Petremand, Yves; Palmieri, Michele; Dubochet, Olivier; Despont, Michel
    We introduce an open, standardized TFLN PICs platform available to PIC designers through a fabless scheme. Several technology steps have been established, providing a reliable process design kit (PDK) in the telecommunication C-band.
  • Item
    An Integrated MEMS Magnetic Gradiometer Rejecting Vibrations and Stray Fields
    (2024-09-09) Gasparin, Enrico; Hoogerwerf, Arno; Bayat, Dara; Spinola Durante, Guido; Petremand, Yves; Tormen, Maurizio; Despont, Michel; Close
    Due to electrification, magnetic sensors are increasingly deployed in magnetically polluted environments. To reject stray fields, differential sensing schemes are typically used. In this paper, we rely instead on a single-point gradiometric sensing scheme based on the force exerted on a magnet, which is directly related to the magnetic field gradient. Unlike the prior art, our concept uses a differential MEMS force-sensing mechanical transducer. The overall sensor rejects not only magnetic stray fields, but also mechanical disturbances such as vibration and gravity. It is the first single-point MEMS gradiometer that can operate unshielded in various orientations. Our prototype realization achieves a noise density of 4 nT/mm/√Hz in a ±300 μT/mm measurement range. We demonstrate its operational effectiveness in a bus-bar current sensing application. Finally, the paper discusses the remaining weaknesses and provides an outlook for MEMS-based magnetic sensors.