This repository contains data tools for exporting from labelme format to a different one. There are also some tools for the dataset in general (sorting based on view, validating annotations).
All python scripts were tested with Python 3.7.10.
As of now, only Windows is supported for starting labelme.
Simply clone this repository. Use update.bat in the tools directory to update to the latest commit.
- The
datadirectory is for your dataset. - The
anno_datadirectory is for your annotation files. - All scripts are in the
toolsdirectory.start_labelme.batstarts labelme (annotation program).update.batfetches and pulls the repository (updates to the latest version).export.pyexports data set annotations to another formats.validate.pyvalidates annotations.datatoolspython module contains internal python code._labelmedirectory contains config for labelme.
You need to have labelme for annotation:
pip install labelmeDouble click start_labelme.bat or use the command line:
start_labelme.bat <optional: specify a picture directory>The default directory is anno_data in the root directory. It is not recommended to change this default.
If you specify the -S flag right after a script, your arguments will be saved into a last_<script>_args.txt file.
For example:
python export.py -S yolo --exec="..\..\darknet\darknet.exe"After that, if you run the script without arguments, it will use the ones in the .txt file.
Currently, export.py supports exporting to YOLOv4 and vehicle attributes formats.
Write into the terminal:
python export.py -h(you might have to change python to python3 if you are not on Windows).
That will show you all the possible formats. For help on export arguments, use the -h or --help flags after a format:
python export.py <format> --helpDefault data directory is .\data. To include data directory in export, prefix it with an ❗. Use the --help flag for information for all arguments.
not done yet 😔
Work in progress.