Machine Learning Application Benchmarking

FPGA   Python   TensorFlow

My helpful screenshot

This image represents a macrophotography of an FPGA (Xilinx ZCU102)

Currently, a lot of satellites orbiting the earth, and a large percentage of them are constantly producing data. This very large amount of information, send down to earth, is stored at institutions like the German Aerospace Center (DLR) as well as companies which are owning a satellite.

The project Machine Learning Application Benchmarking on COTS Inference Processors should demonstrate the ability to apply machine learning algorithms in custom hardware, such as an FPGA Xilinx ZCU102. This project is funded by the European Space Agency (ESA) and is a collaboration between the following parties:

  • Airbus Defence and Space
  • Technical University of Munich
    • Professorship for Big Geospatial Data Management
    • Professorship for Computer Architecture and Parallel Systems
  • OroraTech

The responsibility of us, the professorship of Big Geospatial Data Management is to work on a model zoo for the benchmarking of our FPGA (Xilinx ZCU102). So, we bring knowledge in the fields of Deep Learning, Earth observation, and Computer vision into this project. Additionally, with our knowledge in mobile, embedded and edge computing, we are able to provide small and compressed models to this project.

Published:
We have published so far the following papers:

  1. Ghiglione, M., Serra, V., Raoofy, A., Dax, G., Trinitis, C., Werner, M., Schulz, M., & Furano, G. (2022). Survey of frameworks for inference of neural networks in space data system. DASIA 2022.
  2. Raoofy, A., Dax, G., Serra, V., Ghiglione, M., Werner, M., & Trinitis, C. (2022). Benchmarking and feasibility aspects of machine learning in space systems. Proceedings of the 19th ACM International Conference on Computing Frontiers, 225–226. https://doi.org/10.1145/3528416.3530986
  3. Ghiglione, M., Raoofy, A., Dax, G., Furano, G., Wiest, R., Trinitis, C., Werner, M., Schulz, M., & Langer, M. (2021). Machine Learning Application Benchmark for Satellite On-Board Data Processing. In European Workshop on On-Board Data Processing. https://doi.org/10.5281/zenodo.5520877
  4. Raoofy, A., Dax, G., Ghiglione, M., Langer, M., Trinitis, C., Werner, M., & Schulz, M. (2021). Benchmarking Machine Learning Inference in FPGA-based Accelerated Space Applications. Workshop on Benchmarking Machine Learning Workloads on Emerging Hardware.