Microsoft COCO is both a large-scale dataset and a data annotation format. Since its first release, COCO has emerged as a widely adopted annotation standard for object detection, segmentation, and captioning. Despite Microsoft providing a simple API to interact with the COCO format, performing complex analysis and operations within and across COCO datasets requires ad-hoc implementations by the users.
Our library, COCO Helper, makes working with the COCO format considerably easier. COCO Helper allows to efficiently run complex operations such as statistical analysis and dataset arithmetic using just a few lines of code.
Quick Links:
To install the library you can download the latest wheel package from this repo and install through pip, poetry, or other python package managers.
NB: We are still working to publish the library to public PyPi.
Examples of usage can be found in usage.md and in the notebooks folder.
Contributors are always welcome to help us fix an issue, add tests, improve code and documentation quality. If you'd like to contribute, please see our contributing guide.
Thanks a lot to all of our outstanding contributors!
We would appreciate your citation if you find COCO Helper useful in your research.
You can cite COCO Helper using the following BibTeX entry:
@misc{cocohelper,
title={COCO Helper},
author={TBD},
year={2022},
howpublished={\url{https://github.com/baker-hughes/cocohelper}},
}