This repository contains code for augmentations of images with bounding box annotations. The images are rotated, shifted and zoomed. The bounding boxes are updated accordingly.
The code offers fine-grained control over the augmentation mechanism and integrates nicely with Keras.
Here is an explanation of the augmentation process: https://github.com/lhk/bbox_augmentations/blob/master/showcase.ipynb
Here is demo code on how to integrate this into a machine learning workflow: https://github.com/lhk/bbox_augmentations/blob/master/integrated_demo.ipynb