-
code: complete eval pipeline (7 metrics + per-class + Wilcoxon) + Swin-UNet/TransUNet networks; remove backups/obsolete
-
report: clarify per-class macro vs main-table Dice averaging convention
-
report: add appendix Sensitivity & Precision tables (Spec/ASSD omitted: low discriminative power / redundant)
-
report: drop cross-dataset average columns + overall-ranking table (heterogeneous modalities); per-dataset + significance only
-
report: paper-style layout (transposed methods×datasets, sig-bold, ranking, per-class)
-
aggregate.py: unified-512 intro + per-class + Wilcoxon
-
Upload folder using huggingface_hub
