Edge SpAIce project

Edge SpAIce is an exciting new EU project aiming to deploy AI monitoring onboard Earth Observation satellites for the monitoring of marine plastic litter. The project will be undertaken by a consortium lead by Agenium Space, with CERN, NTUA, and Endurosat. With Agenium Space’s model compression tools, we also aim to demonstrate a pipeline for deployment of generic image segmentation models onboard to perform filtering of relevant images before downlinking them. The CERN team will use the hls4ml tool to convert Deep Neural Networks (DNNs) for Image Segmentation to FPGA firmware for SoCs, targeting high framerate balanced with low power consumption. We expect to develop new features of hls4ml towards supporting significantly larger Image Segmentation models than previously demonstrated. In the second phase we’ll be developing support in hls4ml for NanoXPlore FPGAs, with an effort on backend implementations in hls4ml, strengthening European non-dependence in the sector. All the developments will be made available to the hls4ml users and Fast ML community as we contribute to the Open Source software.

See the Edge SpAIce project website for more information: edgespaice.eu