ruankie / poly2pascal

Convert POLYGON bounding boxes from csv files to xml annotation files in the Pascal VOC format for computer vision

poly2pascal

Convert POLYGON bounding boxes from .csv files to .xml annotation files in the Pascal VOC format for computer vision projects. Read the package documentation here.

Installation

Install poly2pascal with pip

  pip install poly2pascal

Usage

See the example notebook for more details. Here are the essentials of how to use this package:

# import annotator
from poly2pascal.annotations import XMLAnnotator

# create annotator
xmla = XMLAnnotator(
    images_path="path/to/images", 
    csv_file_path="path/to/csv/file.csv", 
    image_name_col="<image_name_column>",
    image_label_col="<object_label_column>", 
    xml_output_path="path/to/xml/annotation/output"
)

# create xml annotation files in Pascal VOC format
xmla.get_all_xml_annotations(img_format=".jpg")

Required Data Format

This package requires a .csv file with the following format of images and polygon annotations. These will be converted to .xml annotation files in the Pascal VOC format:

image_name	object_label	geometry
img_01.jpg	cat	POLYGON ((2478.89 156.77, 2478.89 297.86, 2377.97 297.86, 2377.97 156.77, 2478.89 156.77))
img_02.jpg	cat	POLYGON ((939.81 1221.46, 939.81 1319.68, 715.39 1319.68, 715.39 1221.46, 939.81 1221.46))
img_03.jpg	dog	POLYGON ((1559.7 1558.06, 1559.7 1966.3, 1129.73 1966.3, 1129.73 1558.06, 1559.7 1558.06))

Licence

• MIT

Authors

• @ruankie