javirk / fullweakannotations

Full or Weak annotations?

This is the repository for the paper "Full or Weak annotations? An adaptive strategy for budget-constrained annotation campaigns", presented at CVPR2023. It contains the code to reproduce the experiments presented in the paper.

Links

• Paper
• Project page

How to run

First of all, you need a surface of a dataset. A sample surface has been stored in surfaces/sample_surface.txt. The format is:

run_name, classification share (%), segmentation share (%), IoU, Dice

Then, you will have to create a gp_config file. You can use any gp_config file in the gp_configs folder as a template. Remember to change surface_file parameter to the filename of your surface file.

Finally, you can run the method with the file gp.pyas follows:

python gp.py --config gp_configs/gp_config.txt

If you use this code or the paper, consider citing:

@inproceedings{tejero2023full,
  title={Full or Weak annotations? An adaptive strategy for budget-constrained annotation campaigns},
  author={Tejero, Javier Gamazo and Zinkernagel, Martin S and Wolf, Sebastian and Sznitman, Raphael and Neila, Pablo M{\'a}rquez},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2023}
}