mukopikmin / bounding-box-augmentation

Augment dataset (images and annotations) for object detection

Bounding Box Augmentation

Example of augmentating image with bounding boxes. This code is based on aleju/imgaug.

Augmentable annotation format is Pascal VOC only.

Examples

Above image can be augmented to the following images.

Example 1	Example 2	Example 3	Example 4

Requirements

• Python3
• Pipenv

Usage

Prepare virtualenv with pipenv.

pipenv install

Put files(Images and Annotations) to augment in input/. (Do not create sub directories.)

Run script.

pipenv run python augment.py

Augmented images and annotations are generated in output/. If there is no bounding boxes in the image, Annotation files (*.xml) are moved to empty/.