Update README

README

@@ -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).