👉 Convert object detection dataset format 👈

Dataset types

PASCAL VOC: Pascal voc dataset have a xml file for each image.

YOLO: YOLO dataset have a txt file for each image.

COCO: COCO dataset have a json file for each image.

Current support format

Currently, the following formats are supported:

from	to	implemented
PASCAL VOC	YOLO(TXT files)	Yes
YOLO	PASCAL VOC (XML files)	Yes

Upcoming support format

from	to	Issue/PR(if any)
PASCAL VOC	COCO (JSON files)	No
PASCAL VOC	TFRecord (TFRecord files)	No
COCO	PASCAL VOC (XML files)	No
COCO	YOLO (TXT files)	No
COCO	TFRecord (TFRecord files)	No
YOLO	COCO (JSON files)	No
YOLO	TFRecord (TFRecord files)	No

Installation

Installation from source code

git clone https://github.com/codePerfectPlus/dataset-convertor/

cd dataset-convertor
python -m venv venv

source venv/bin/activate
pip install requirements.txt

Installation from PyPI

pip install dataset-convertor

Usage

convert annotations from one format to another format.

dataset formatting example:

- data/pascal_voc/JPEGImages/*.jpg
- data/pascal_voc/Annotations/*.xml

- data/yolo5/JPEGImages/*.jpg
- data/yolo5/labels/*.txt

Pascal VOC(xml) to yolo(txt)

from convert import Convertor

con = Convertor(input_folder='/home/user/data/pascal_voc', output_folder='/home/user/data/yolo5')
con.voc2yolo()

from yolo(txt) to Pascal VOC(xml)

from convert import Convertor
con = Convertor(input_folder='/home/user/data/yolo5', output_folder='/home/user/data/pascal_voc')
con.yolo2voc()

Contributing

create an issue/PR if any format is missing.Open-source contribution is welcome.check the contributing guide for details.

Reference

• PASCAL VOC - http://host.robots.ox.ac.uk/pascal/VOC/
• COCO - http://cocodataset.org/
• YOLO9000 - https://arxiv.org/abs/1612.08242
• YOLO4 - https://arxiv.org/abs/2004.10934v1

License

• MIT License

Authors

• CodePerfectPlus