Butui Hu
Butui Hu
Maybe I should ask this question on Hugo's [discourse](https://discourse.gohugo.io). For now, I use [diagrams](https://github.com/francoislaberge/diagrams) to generate the flowchart as svg file, and then insert in my post.
Sorry, this code is buggy now. You could try another implement.
Emm, a typo? I don't see any differences.
I'm not sure. I don't do classification. I only do the segmentation, the classification result is calculated from the segmentation result based on the Gleason system.
You could merge these classes. I have a long time not working on this. You could try your way.
You could always check what Gleason system is.
Everything you need is in the code, https://github.com/hubutui/Gleason/blob/master/train.py#L205
The valid label value is 0-6, but there is no image in trainset with label 2. To simplify, you could remap the label or just leave it unchanged. There is...
`undecidedlabel` is an arg required by `sitk.MultiLabelSTAPLE`, you could check it [here](https://itk.org/SimpleITKDoxygen/html/classitk_1_1simple_1_1MultiLabelSTAPLEImageFilter.html). `batch_size=2`, 200 images for training, and 44 for validation.
@phuongchi911 I don't understand your question. Did you download the dataset? The masks, which are saved as images, store label value as pixel intensity, usually ranging from 0 to C-1...