Dataset format converter for object detection

Deep learning field for image processing has many dataset format. For object detection task, MSCOCO, YOLO, PascalVOC and so on are frequently used and sometimes we have to convert our dataset annotation files to another. To deal with this, I creat dataset format converter for object detection. Supported dataset formats are below.

• MSCOCO
• YOLO
• PascalVOC
• KITTI

• Python >= 3.7

Install python>=3.7 and run commands below in your venv or something.

pip install -U pip setuptools build
python -m build
pip install dist/objdet_converter-1.0.0-py3-none-any.whl

Call 'convert_format' function with some args.

from objdet_converter.convert import convert_format

convert_format(
    src_format="coco",
    dst_format="yolo",
    src_path="./coco.json",
    dst_path="./yolo_output",
    class_txt_path="",
)

Sample code is available here. Detail descriptions are here

If args of 'class_txt_path' is empty, scan all annotation files and create class list automaticaly(alphabetical order).

• MSCOCO
• YOLO
• PascalVOC
• KITTI
• Custom Dataset Implementation