qurator-spk/sbb_pixelwise_segmentation
1
0
Fork 0
mirror of https://github.com/qurator-spk/sbb_pixelwise_segmentation.git synced 2025-06-09 11:50:04 +02:00
Code Issues Releases Wiki Activity

Update README

Browse source
This commit is contained in:
Rezanezhad, Vahid 2019-12-05 14:08:08 +01:00
parent cd1990d4f9
commit a216dccfcf
1 changed files with 8 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

13
README
View file

@ -4,17 +4,20 @@ how to train:
format of ground truth:
Lables for each pixel is identified by a number . So if you have a binary case n_classes should be set to 2 and labels should be 0 and 1 for each class and pixel.
In the case of multiclass just set n_classes to the number of classes you have and the try to produce the labels by pixels from 0 , 1 ,2 .., n_classes-1.
Lables for each pixel is identified by a number . So if you have a binary case n_classes should be set to 2 and
labels should be 0 and 1 for each class and pixel.
In the case of multiclass just set n_classes to the number of classes you have and the try to produce the labels
by pixels set from 0 , 1 ,2 .., n_classes-1.
The labels format should be png.
If you have an image label for binary case it should look like this:
Label: [ [[1 0 0 1], [1 0 0 1] ,[1 0 0 1]], [[1 0 0 1], [1 0 0 1] ,[1 0 0 1]] ,[[1 0 0 1], [1 0 0 1] ,[1 0 0 1]] ] this means that you have an image by 3*4*3 and pixel[0,0] belongs to class 1 and pixel[0,1] to class 0.
Label: [ [[1 0 0 1], [1 0 0 1] ,[1 0 0 1]], [[1 0 0 1], [1 0 0 1] ,[1 0 0 1]] ,[[1 0 0 1], [1 0 0 1] ,[1 0 0 1]] ]
this means that you have an image by 3*4*3 and pixel[0,0] belongs to class 1 and pixel[0,1] to class 0.
traing , evaluation and output:
training , evaluation and output:
train and evaluation folder should have subfolder of images and labels.
And output folder should be free folder which the output model will be written there.
And output folder should be empty folder which the output model will be written there.
patches: