Iterative Refinement Strategy for Automated Data Labeling

The code has been successfully tested on Ubuntu 22.04.

Create Conda Environment

$ conda create -n yolov8 python=3.10 -y
$ conda activate yolov8
$ https://github.com/wish44165/iAutolabeling
$ cd iAutolabeling/
$ pip install ultralytics

Commands

$ for i in `seq 0 9`; do python main.py --curr_iter ${i} | tee iterLog${i}.txt; done

Folder Structure

Initial

ICME2024/
├── datasets/
    └── v0/
        ├── images/
            ├── train/
            └── val/
        └── labels/
            ├── train/
            └── val/
└── src/
    └── iAutolabeling/
        ├── facial.yaml
        └── main.py

After executed

ICME2024/
├── datasets/
    ├── v0/
        ├── images/
            ├── train/
            └── val/
        └── labels/
            ├── train/
            └── val/
    └── v1/, v2/, ...
├── src/
    └── iAutolabeling/
        ├── facial.yaml
        ├── main.py
        ├── facial_v1.yaml, facial_v2.yaml, ...
        ├── iterLog0.txt, iterLog1.txt, ...
        └── runs/
            └── facial/
                ├── train/, train2/, ...
                └── predict/, predict2/, predict3/, predict4/, ...

Acknowledgments

• Ultralytics
• Model Prediction with Ultralytics YOLO
• Problem (solution)

Citation

@misc{chen2024iterative,
      title={Iterative Refinement Strategy for Automated Data Labeling: Facial Landmark Diagnosis in Medical Imaging}, 
      author={Yu-Hsi Chen},
      year={2024},
      eprint={2404.05348},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}