- User Since
- Aug 1 2016, 12:10 PM (106 w, 2 d)
Jul 2 2018
Oversampling foreground makes the dice believe the foreground is more frequent than it actually is. Therefore, it does not punish false positives hard enough. I therefore tried to balance the dice by doing the following:
Similar to Jakobs segmentation method. 3 UNets, each trained with a different perspective. Little improvement on Hippocampus, no difference on BraTS, but worse performance on other datasets -> Max be due to anisotropy
Dec 20 2017
Oct 24 2017
Feb 3 2017
Jan 23 2017
Bug is not in the plugin (oops).
The plugin calls the GetStatistics method twice, but should not trigger the calculation 2 but instead simply return the already computed statistics at the second call.
The bug is caused by improper handling of modified() functionality in the image statistics calculator and in the mask generators. Whenever the time step is changed, both will be modified which is noticed by the image statistics calculator and will lead to recomputation of statistics.