Notebooks and code samples to help you use the PyLabel Python package and the PyLabeler Jupyter-based labeling tool.

Annotation Format Conversion

Use PyLabel to translate bounding box annotations between different formats-for example, from coco to yolo.

• coco2voc.ipynb
• coco2yolov5.ipynb
• voc2coco.ipynb
• yolo2coco.ipynb
• yolo2voc.ipynb
• yolo_with_yaml_importer.ipynb

PyLabeler

PyLabeler is a Jupyter-based labeling tool that you can use to annotate images and edit bounding box annotations within a Jupyter notebook.

• label_new_dataset.ipynb
  This notebook is a labeling tool that can be used to annotate image datasets with bounding boxes, automatically suggest bounding boxes using an object detection model, and save the annotations in YOCO, COCO, or VOC format.
• yolo2pylabeler.ipynb
  This notebook uses PyLabeler to edit an existing dataset of Yolo annotations and save the new annotations back to Yolo format.

Integrations with Other Tools

PyLabel can help you use other tools that take bounding box annotations as an input or output. PyLabel stores annotations as a Pandas dataframe, which you can access directly to support your particular use case.

• albumentations.ipynb
  If you don't have enough images to train a model well, you can use image augmenation to create more samples for training and validation. Albumentations is a popular open-source library for creating additional, augmented images as well as the annotations for those images.

• azure_custom_vision.ipynb
  Using PyLabel you can import existing labels in COCO, YOLOv5, or VOC format and then upload the dataset to Azure Custom Vision.