Configuration

Configuration

Copy config.txt and change the values you want.

dataset

• in_path : path for input image
• gd_path : path for ground truth image
• train : number of image to use for training
• valid : number of image to use for validation
• test : number of image to use for prediction

Precondition : train + valid + test <= number of images
The image are sorted in alphabetical order.

train

• patch_size_x : patch size (x axis)
• patch_size_y : patch size (y axis)
• patch_size_z : patch size (z axis)
• batch_size : size of batch
• steps_per_epoch : steps per epoch
• epochs : number of epochs
• logs_path = path for logs (logs folder is used for tensorboard logs, training logs and saved models)

If all patch_size are equal to the image size, the training will be done on the full images.