Generative inter-class transformations for imbalanced data weather classification

This paper presents an evaluation of how data augmentation and inter-class transformations can be used to synthesize training in low-data scenarios for single-image weather classification. In such scenarios, augmentations is a critical component, but there limit much improvements gained using classical strategies. Generative adversarial networks (GAN) have been demonstrated generate impressive results, also successful as tool augmentation, mostly images limited diversity, medical applications. We investigate the possibilities generative balancing small classification dataset, where one class has reduced number images. compare intra-class by means well noise-to-image GANs, interclass from another are transformed underrepresented class. The results show that it possible take advantage GANs balance dataset opens up future work on GAN-based both diverse scarce.