HeadAttribute

A label tool for annotating human head (belonging to rigid objects) pose in 2D image with standard 3D cube

Introduction

• this project is based the popular image annotation tool https://github.com/HumanSignal/labelImg
• this is a simple demo tool for answering my issue https://github.com/cvat-ai/cvat/issues/3387 under the CVAT project
• it can be used for annotating multiple human head locations and 3D poses from only 2D images
• the current labelImg.exe file can only be executed on Windows, not supporting Linux/Ubuntu/Mac
• please download the released file labelImg.exe and put it under the subfolder ./dist/ for running
• possible local workdir is C:\Users\zhy19\AppData\Local\Temp\_MEI1731402 when running labelImg.exe
• you can change the static 3D head model ./dist/data/Female3DHead.obj into your ideal one

Running Advice

If you just want to use this annotation tool for labeling human head poses, thes just run labelImg.exe and follow user manuals of original labelImg. Below is an example video of using the tool.

https://github.com/user-attachments/assets/4b413102-c61e-46f6-b027-ce595c14acee

Labeling steps: Open Dir --> Change Save Dir --> Next Image --> Begin Annotating with Careful Adjusting --> Save

Compiling Advice

• Follow labelImg_v6.spec to create a new python env using conda (e.g., py37_headpose).
• Also remember install many third dependencies listed in the file labelImg_v6.spec
• You can develop your ideal version by modifying .py files under the foler ./libs
• Debug your modified version by running the file run.bat for callbacking labelImg.py
• Install pyinstaller for packaging all source code and dependencies into one .exe file

References

Citation of labelImg: Tzutalin. LabelImg. Git code (2015). https://github.com/tzutalin/labelImg

Citation of HeadAttribute: This project is also based on my previous work for simultaneous head detection and pose estimation

@article{zhou2022intuitive,
  title={An Intuitive and unconstrained 2D cube representation for simultaneous head detection and pose estimation},
  author={Zhou, Huayi and Jiang, Fei and Xiong, Lili and Lu, Hongtao},
  journal={arXiv preprint arXiv:2212.03623},
  year={2022}
}