Skip to main content
. Author manuscript; available in PMC: 2024 Oct 2.
Published in final edited form as: IEEE Trans Med Imaging. 2023 Oct 2;42(10):2832–2841. doi: 10.1109/TMI.2023.3266137

Fig. 1.

Fig. 1.

The architecture of Min-Max Similarity. {ℱ1,2},{𝒞1,𝒞2}, and {𝒫1,𝒫2} denote segmentation networks, classifiers and projectors, respectively. Here, classifiers extract features to build all-negative pairs, and projectors project unlabeled predictions to high dimension features for pixel-wise consistency measurement.