- polygon-format can be explicitly enforced rather than auto-detected using the --use-polygon-format option when reading annotations using from-yolo-od
- supports polygon format now (index + pairs of normalized x/y coordinates)
- YOLOODReader now skips annotations if it fails to locate associated image
- the read_labels_file method of FromYOLOOD now strips leading/trailing whitespaces from the labels.
- Initial release.