This is a library to evaluate click-based interactive segmentation models. isegeval could evaluate the number of click (NoC) performance of the given model on the given dataset.
You could evaluate your model as follows. See notebooks for more detail.
from isegeval import evaluate
from isegeval.core import ModelFactory
# Each item is the tuple of an image and its correspoinding ground truth mask.
dataset: Sequence[tuple[np.ndarray, np.ndarray]] = YourDataset()
# A factory of your model that you want to evaluate. The factory should implement get_model method.
model_factory: ModelFactory = YourModelFactory()
evaluate(dataset, model_factory)pip install isegeval