Ensembling it with the tumor predictions of another network (i.e. the sigmoid network) can boost the horrible tumor segmentation performance.
The sequence of PR would go from #25 -> #6 -> I/O fixes for binary tumor segmentation.
Side note: This needs to be done for both mean_std_norm and w/o it
Ensembling it with the tumor predictions of another network (i.e. the sigmoid network) can boost the horrible tumor segmentation performance. The sequence of PR would go from #25 -> #6 -> I/O fixes for binary tumor segmentation. Side note: This needs to be done for both mean_std_norm and w/o it