Active Segmentation¶
An active learning benchmarking framework for scientists seeking comparability and reproducibility.
active segmentation provides a flexible and expressive API for evaluating active learning strategies for
medical image segmentation.
With active segmentation, you can:
Define own segmentation models by implementing
models.pytorch_model.PytorchModel(), or use pre-configured models like the U-Net.Add own datasets by implementing
datasets.data_module.ActiveLearningDataModule(), or use already added datasets like the medical segmentation decathlon.Evaluate active learning query strategies<query_strategies, or define and test new strategies by implementing
query_strategies.query_strategy.QueryStrategy().Train two-dimensional or three-dimensional segmentation models.
Run fully reproducible experiments, with seeded random processes and only deterministic operations.
Track various metrics with Weights and Biases.
Issues¶
Submit issues, feature requests or bugfixes on github.
How to Cite¶
If you use active segmentation in the context of academic or industry research, please
consider citing the paper.
Paper¶
@misc{https://doi.org/10.48550/arxiv.2207.00845,
doi = {10.48550/ARXIV.2207.00845},
url = {https://arxiv.org/abs/2207.00845},
author = {Burmeister, Josafat-Mattias and Fernandez Rosas, Marcel and Hagemann, Johannes and Kordt, Jonas and Blum, Jasper and Shabo, Simon and Bergner, Benjamin and Lippert, Christoph},
keywords = {Computer Vision and Pattern Recognition (cs.CV), Artificial Intelligence (cs.AI), Machine Learning (cs.LG), FOS: Computer and information sciences, FOS: Computer and information sciences, I.4.6; I.2.10; J.3},
title = {Less Is More: A Comparison of Active Learning Strategies for 3D Medical Image Segmentation},
publisher = {arXiv},
year = {2022},
copyright = {arXiv.org perpetual, non-exclusive license}
}
License¶
active segmentation is licensed under the AGPL-3.0 license.