AI data augmentation through simulation
MSTAR dataset is a publicy available dataset containing SAR imagettes of a set of military targets.
It was collected in 1995 at the Redstone Arsenal, Huntsville, AL by the Sandia National Laboratory using an X-band SAR sensor.
The dataset has become a standard-de-facto validation input for SAR Automatic Target Recognition algorithms
AI data augmentation
In our case study, we leveraged the MSTAR dataset for two different purposes:
Simulator calibration
Training our AI models on MSTAR real data and then test them on SimRAD simulated data
Data augmentation
Training our AI models on SimRAD simulated data and test them on real MSTAR data
Sample results
