Generative Adversarial Networks for Vehicle Transformation in Overhead and Satellite Imagery
Michael J. Reale, Preston Nichols, Ethan Schneider, Morgan A. Bishop, Maria Cornacchia · 2022
The ability to create and detect synthetic video is becoming critically important to scene understanding. Techniques for synthetic manipulation and augmentation of data increases diversity within available datasets, while not requiring laborious labeling efforts. That is, the ability to create synthetic video can enable augmentation of small realistic datasets on which to further train Artificial Intelligence and Machine Learning (AI/ML) algorithms. Thus, it may be desirable to convert images of vehicles in satellite and overhead imagery to other varieties of vehicles. In this work, we leverage generative adversarial networks to transform cars into trucks (and vice versa). We leverage an attention-based masking approach that assists the network in transformation of the object and not background. In addition, we demonstrate the benefits of numerous data augmentation procedures, including presenting a new artificial dataset of vehicles from an aerial perspective and introducing novel augmentation techniques appropriate for our network architectures. Experiments are conducted for this unique application on both real and artificial datasets with state-of-the-art results.