Fig 2 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Source publication
We introduce unORANIC, an unsupervised approach that uses an adapted loss function to drive the orthogonalization of anatomy and image-characteristic features. The method is versatile for diverse modalities and tasks, as it does not require domain knowledge, paired data samples, or labels. During test time unORANIC is applied to potentially corrupt...
Context in source publication
Context 1
... entire training pipeline is depicted in Fig. 2. Both encoders consist of four identical blocks in total, where each block comprises a residual block followed by a downsampling block. Each residual block itself consists of twice a convolution followed by batch normalization and the Leaky ReLU activation function. The downsampling blocks use a set of a strided convolution, a batch ...