FURI | Fall 2020
A Benchmarking Framework for Data-Driven Compressive Sensing
The objective of the research is to develop a benchmarking framework with a unified API and benchmarks to allow researchers in this field to quantitatively evaluate new algorithms. To achieve fair and extensive comparisons, all data-driven methods are implemented in PyTorch, and are compared to state-of-the-art model-based methods that are implemented in MATLAB, with a variety of dataset and parameter setups. The CSGM and CSGAN algorithms were reimplemented in PyTorch as part of this framework, based on the provided TensorFlow implementations. The fair evaluation of these new algorithms could ultimately lead to improvements in the practical applications of Compressive Sensing.
Hometown: Budapest, Pest megye, Hungary
Graduation date: Fall 2020