Aata augmentation for bounding boxes

We adapted image augmentation techniques to object detection tasks. We implemented rotation, width and height shift, and scaling of images as well as bounding boxes using OpenCV's affine transformation features. We assumed only one object (bounding box) per image.