Filters

Settings

Now showing1 - 2 of 2
  • Some of the metrics are blocked by your 
    consent settings
    Item type:Publication, Access status: without files ,
    EdgeBoost: confidence boosting for resource constrained inference via selective offloading
    (IEEE, 2024-04)
    Said, Naina
    ;
    Landsiedel, Olaf
    This paper introduces EdgeBoost, a selective input offloading system designed to overcome the challenges of limited computational resources on edge devices. EdgeBoost trains and calibrates a lightweight model for deployment on the edge and, in addition, deploys a large, complex model on the cloud. During inference, the edge model makes initial predictions for input samples, and if the confidence of the prediction is low, the sample is sent to the cloud model for further processing otherwise, we accept the local prediction. Due to careful calibration, EdgeBoost reduces the communication cost by 55%, 29% and 20% for CIFAR-100, ImageNet and Stanford Cars datasets, respectively, when compared to an cloud-only solution while achieving on par classification accuracy. Finally, EdgeBoost also achieves comparable accuracy to the state-of-the-art routing-based methods without the need for hosting the router on the edge.
    Publicationtype: Conference Paper
      37
  • Some of the metrics are blocked by your 
    consent settings
    Item type:Publication, Access status: without files ,
    ISFD : Efficient and Fault-tolerant In-System-Failure-Detection for LP FPGA-based Smart-Sensors in Space Expeditions
    (IEEE, 2024-04)
    Rahman, Kazi Mohammad Abidur
    ;
    Dirkes, Timo
    ;
    Delfs, Bjoern
    ;
    Wyrwoll, Vanessa
    ;
    Kulau, Ulf
    The health of astronauts is a top priority in manned space missions, and space conditions can have a major impact on the physiology of the human body. This is why extensive health monitoring is essential, especially for future missions beyond low Earth orbit (LEO), where astronauts will be increasingly exposed to higher radiation doses. To address these challenges, the AuRelia project has developed space-capable smart sensors that are designed to be failure-tolerant and autonomous. These sensors employ an innovative In-System-Failure-Detection (ISFD) technique that caters to the increasing need for Single Event Effect (SEE)-induced failure tolerance in resource-constrained and ultra-low power SRAM-based FPGAs used in miniaturized smart sensors. The ISFD technique integrates a hardware-based radiation-tolerant Test Core (using only 77 LUTs) with a software-based Test Controller, resulting in a lightweight yet efficient solution for failure detection. What's more, the adaptability of ISFD facilitates seamless integration into existing firmware and software, offering developers a swift implementation process. ISFD complements traditional blind scrubbing methods, enhancing sensor data reliability, and has been validated with radiation up to 70Krad.
    Publicationtype: Conference Paper
      12