Usage
Select the DL4Mic_Toolset from the toolsets button (>>) of ImageJ:
Open the about dialog.
Open the help page.
Select the network you want to train or apply from this menu-button.
Display information about the selected network.
The upper part of the menu allows to set the parameter of the selected network. The lower part allows to configure dl4mic.
-
user parameters A group of parameters that you might want to modify.
-
advanced parameters Only modify these parameters if you know what you are doing
-
internal network parameters Only modify these parameters if you are an expert for artificial neural networks.
-
python interpreter Set the path to the python executable if it has not been found automatically.
-
install deep-learning env. Install the conda-env in which the python code will be executed. You need to do this one time before using a network, see Installation.
Run the training of the network, using the images and parameters configured in the user parameters. When the training is finished, a plot of the loss and of the validation loss will be displayed.
Evaluate the network. Different metrics, depending on the given network, will be calculated and displayed.
Apply the trained network to your images. If a single image or a stack is open, it will be used as input data and the result will likewise be opened as an image or a stack. Otherwise the tool will ask for a folder containing the input images and a folder into which the result images will be saved. Depending on the network, more than one result image might be created per input image, for example a probability map and a binary mask.