Data Augmentation from Proficient Pre-Training of Robust Generative Adversarial Networks (DAPPER GAN)

Military analysts are responsible for a variety of critical tasks, such as maintaining situational awareness and performing rapid, high-consequence decision making based on electro-optical, infrared, and full motion video data sources. Recent advances in machine learning (ML) and computer vision tools promise to enhance analyst workflows and reduce cognitive burden, but their successful adoption poses several challenges: (1) analytics must support objects of interest with sparse data; (2) ML tools are sensitive to their training sets and typically require cumbersome, labor-intensive training and validation processes; and (3) ML tools can be extremely “black box,” offering analysts little to no insight into the decision-making process or how new data might affect prediction performance. To address these challenges, we propose to design and demonstrate the feasibility of Data Augmentation from Proficient Pre-Training of Robust Generative Adversarial Networks (DAPPER GAN), a ML analytics support tool that automatically generates and maintains data sets to validate and enhance the performance of machine learning algorithms.