qurator-spk/sbb_pixelwise_segmentation
1
0
Fork 0
mirror of https://github.com/qurator-spk/sbb_pixelwise_segmentation.git synced 2025-06-07 19:05:24 +02:00
Code Issues Releases Wiki Activity

Update README

Browse source
This commit is contained in:
vahidrezanezhad 2019-12-10 11:57:37 +01:00 • committed by GitHub
parent f69d4459cb
commit 0cddfff04f
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 16 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

22
README.md
View file

@ -4,16 +4,21 @@
# Ground truth format # Ground truth format
Lables for each pixel is identified by a number . So if you have a binary case n_classes should be set to 2 and 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. 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 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. by pixels set from 0 , 1 ,2 .., n_classes-1.
The labels format should be png. The labels format should be png.
If you have an image label for binary case it should look like this: 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]] ] Label: [ [[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. [[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.
# Training , evaluation and output # Training , evaluation and output
train and evaluation folder should have subfolder of images and labels. train and evaluation folder should have subfolder of images and labels.
@ -21,6 +26,11 @@
# Patches # Patches
if you want to train your model with patches, the height and width of patches should be defined and also number of if you want to train your model with patches, the height and width of
patches should be defined and also number of
batchs (how many patches should be seen by model by each iteration). batchs (how many patches should be seen by model by each iteration).
In the case that model should see the image once, like page extraction, the patches should be set to false. In the case that model should see the image once, like page extraction,
the patches should be set to false.
# Pretrained encoder
Download weights from this limk and add it to pretrained_model folder.
https://file.spk-berlin.de:8443/pretrained_encoder/