Generative Data Augmentation for Vehicle Detection in Aerial Images

Scarcity of training data is one of the prominent problemsfor deep networks which require large amounts data. Data augmentationis a widely used method to increase the number of training samples andtheir variations. In this paper, we focus on improving vehicle detectionperformance in aerial images and propose a generative augmentationmethod which does not need any extra supervision than the boundingbox annotations of the vehicle objects in the training dataset. The pro-posed method increases the performance of vehicle detection by allowingdetectors to be trained with higher number of instances, especially whenthere are limited number of training instances. The proposed method isgeneric in the sense that it can be integrated with different generators.The experiments show that the method increases the Average Precisionby up to 25.2% and 25.7% when integrated with Pluralistic and DeepF
Citation Formats
H. Kumdakçı, C. Öngün, and A. Temizel, “Generative Data Augmentation for Vehicle Detection in Aerial Images,” presented at the Workshop on Analysis of Aerial Motion Imagery (WAAMI 2020) in conjunction with25th International Conference on Pattern Recognition (ICPR 2020), 2021, Accessed: 00, 2021. [Online]. Available: