Camera localization predicts the camera pose from a query image. There are two types of deep learning-based camera localization methods: image-based and structure-based. Previous works have shown that data augmentation can improve the performance of image-based methods, but there are no research studies on the structure-based method with data augmentation technique. In this paper, we propose a new pose augmentation procedure that can further improve the performance of the deep structure-based camera localization method, especially under few-shot settings. We investigate different inpainting and rendering strategies and compare their performance with pose augmentation. In addition, we propose a confidence-based sampling scheme that drastically reduces the computation time while maintaining high pose estimation accuracy.