how to use DeepLearning code to make image segmentation

The DeepLearning codes are in the site https://github.com/lisa-lab/DeepLearningTutorials. I want segment a cell image with the codes,but I cann't know how to invoke them. The problems are list as follows: 1. Which model's effect is the best to segment? 2. The each element value of the matrixs train_x,train_y,test_x is only 0 or 1 ? If only the grayscale values of the cell image are introduce, how to make the grayscale values correspond to the element values of the matrix train_x,train_y,test_x? 3. User can draw the outline of objective and backgroud on the cell image to set their grayscale value and label into train_x,train_y, what can I do? 4. The output is the matrix of probabilities, how to use the probability? Thank you!