Few-Shot Aircraft Image Data Augmentation Based on Generative Adversarial Networks

In this paper, a generative adversarial network (GAN) approach is employed to generate aircraft images through various models such as DCGAN, WGAN-GP and StyleGAN3, aiming to augment aircraft image datasets. Based on the FGVCAircraft dataset, comparative experiments are conducted to evaluate the quality of the generated images, and the best model is selected to expand the base dataset to form an enhanced dataset, which is further validated in the YOLO v12 target detection model in terms of its effectiveness in detection accuracy enhancement.