English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

One of the major bottlenecks in high-resolution Earth Observation (EO) spacesystems is the downlink between the satellite and the ground. Due to hardwarelimitations, on-board power limitations or ground-station operation costs,there is a strong need to reduce the amount of data transmitted. Variousprocessing methods can be used to compress the data. One of them is the use ofon-board deep learning to extract relevant information in the data. However,most ground-based deep neural network parameters and computations are performedusing single-precision floating-point arithmetic, which is not adapted to thecontext of on-board processing. We propose to rely on quantized neural networksand study how to combine low precision (mini) floating-point arithmetic with aQuantization-Aware Training methodology. We evaluate our approach with asemantic segmentation task for ship detection using satellite images from theAirbus Ship dataset. Our results show that 6-bit floating-point quantizationfor both weights and activations can compete with single-precision withoutsignificant accuracy degradation. Using a Thin U-Net 32 model, only a 0.3%accuracy degradation is observed with 6-bit minifloat quantization (a 6-bitequivalent integer-based approach leads to a 0.5% degradation). An initialhardware study also confirms the potential impact of such low-precisionfloating-point designs, but further investigation at the scale of a fullinference accelerator is needed before concluding whether they are relevant ina practical on-board scenario.

Low-Precision Floating-Point for Efficient On-Board Deep Neural Network Processing