WP4: Testbed Software Interface and Validation Tests

---

WP4 focuses on developing a software framework to port machine learning (ML) models onto FPGA prototypes created in WP3. This includes creating a virtual interface layer to connect high-level frameworks like TensorFlow, PyTorch, and ONNX with hardware designs. Two solutions will be explored: application-dependent FPGA designs and application-independent designs, like a soft-GPU, offering more flexibility for ML programmers.

In addition to the software framework, WP4 will define and execute a validation plan for testing the performance of selected ML algorithms on the FPGA implementations. Functional and performance tests will be conducted using a pre-selected dataset to evaluate the effectiveness of both the application-dependent FPGA and soft-GPU solutions in various space mission scenarios. The results will guide the final selection of the most suitable solutions for in-flight use.