Stress-testing biomedical vision models with RadEdit: A synthetic data approach for robust model deployment
This paper has been accepted at the 18th European Conference on Computer Vision (ECCV 2024) (opens in new tab), the premier gathering on computer vision and machine learning.
