Rohde & Schwarz Challange at TUM Hackathon

Team µ~ (MicroWave)

The Rohde & Schwarz QAR50

We try to estimate the Reconstruction of the QAR50 and improve the measurement quality.

We used simple shaped objects for the calibration of our algorithm.

Automated Label Generation

Estimation of the reconstruction filter by minimazing the MSError in the Fourierdomain of the magnitude

Final Reconstruction Filter. High Frequenceys are amplified.

Results

How to use this repo

The training images are in: https://github.com/corner100/MikroWave/tree/main/trainset

The Testset is in: https://github.com/corner100/MikroWave/tree/main/testset/images

Run https://github.com/corner100/MikroWave/tree/main/net.py to reproduce our results.

You just need to generate a reconstruction filter once and use it for every arbitrary objects.

To generate the GT Data run https://github.com/corner100/MikroWave/blob/main/generate_gt_from_measurement_coin.py You can use this script for coins, square, hexagon, and the c shaped object.