Synthetic Data Generation with Active Learning Workflows for Radar
, Rendered.AI
Developers face many challenges when integrating simulated or synthetic data into their workflows. These include the physical realism of the simulations, domain transfer, and generating data with sufficient diversity or entropy. We'll show how a closed loop workflow can help overcome these challenges using physics-based simulation, parallel model generation, and active learning techniques. We'll also highlight how simulated data can be used to identify and address failure modes in model performance. This workflow will be demonstrated in the difficult environment of processing radar imagery.