Update README

This commit is contained in:
Rezanezhad, Vahid 2019-12-05 16:11:31 +01:00
parent 036e2e9de5
commit f7a5a57245

8
README
View file

@ -1,8 +1,8 @@
how to train:
# Train
just run: python train.py with config_params.json
format of ground truth:
# 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
labels should be 0 and 1 for each class and pixel.
@ -15,11 +15,11 @@ format of ground truth:
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.
training , evaluation and output:
# Training , evaluation and output
train and evaluation folder should have subfolder of images and labels.
And output folder should be empty folder which the output model will be written there.
patches:
# Patches
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).