• U-net allowed for ability to customize exact architecture
• nnU-Net: A more focused approach to U-net, focus on aspects that make out the perfor- mance and generalizability of a method
• Robust and Self adapting framework??

Problem

• For sota performance, requirement of modification of arch of unet
• arch validation on (a lot of times) a single dataset.
• hard to validate superiority of arch, for the general case (across multiple datasets).

- arch tweeks can be shown to imporve performance for an unoptimized networks, but according to the paper, are unable to make results better for fully optimized one and push sota.

Medical Segmentation Decathlon

• participants asked to create a segmentation algorithm that generalizes across 10 datasets corresponding to different entities of the human body
• allowed to adapt to the dataset but must do so in an automated manner. 1) a development phase in which participants are given access to 7 datasets to optimize their approach on and, using their final and thus frozen method, must submit segmentations for the corresponding 7 held-out test sets. 2) a second phase to evaluate the same exact method on 3 previously undisclosed datasets.