From a216dccfcfa1c9541508a766e2f5dee21e7065d1 Mon Sep 17 00:00:00 2001 From: "Rezanezhad, Vahid" Date: Thu, 5 Dec 2019 14:08:08 +0100 Subject: [PATCH] Update README --- README | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) diff --git a/README b/README index 7d8d790..8d478bd 100644 --- a/README +++ b/README @@ -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: