Important! The original annotations should be in coco format.
To run the benchmark, run the following:
python noise_annotations.py /path/to/annotations --benchmark {easy, medium, hard} (choose the benchmark level) --seed 1
For example:
python noise_annotations.py /path/to/annotations --benchmark easy --seed 1
To run a custom noise method, run the following:
python noise_annotations.py /path/to/annotations --method_name… See the full description on the dataset page: https://huggingface.co/datasets/eden500/COCO-N.
