Skip to content

Classes and plates

A model starts as a list of classes, each a list of plates. Every ticked colony of a class's plates is a training example of that class. Work on the Training Set tab, in Training Mode.

The Training Set toolbar

Button Tooltip Does
Stack of plates, + Add Class With the switch on files, asks for a name. With the switch on the database, opens the Class dialog
Stack of plates, - Remove Selected Class Removes the class selected in the tree. If there is a trained SVM, it asks first: The SVM Model will be destroyed. Are You Sure?
Plate, + Add Plate to the Selected Class Opens .plt files, or with the switch on the database the Database Control Panel
Sheet, green arrow Import Directory into the Selected Class Opens a plate folder
Plate, - Remove Active Plate Removes the active plate from its class
Mortarboard Train the Classifier Trains the SVM
Red cross Remove All Classes Empties the model. If it has not been saved, it asks whether to save it first
Floppy disk Save Model Saves the model
Folder Load Model Loads a model

Making a class

Click Add Class and type a name. Baclan suggests a default one. Buttons that add a plate work on the class selected in the tree, or the class of the plate selected in it. With nothing selected they say Select Class Pls.

To rename a class, click its name in the tree twice, slowly, and type.

With a database

With the switch on the database, Add Class opens the Class dialog. Choose a class from the list, with its plates shown beside it, and press Open to load it; or type a name under New Class and press Create New Class. Save Class on the top toolbar saves a class back to the database.

Many classes at once

Right-click Add Class to load a whole list from a text file: a class name, its plate folders one per line, a blank line, the next class name, and so on.

Listeria monocytogenes
D:\Plates\LM001
D:\Plates\LM002

Salmonella
D:\Plates\SAL015
D:\Plates\SAL049

Choosing colonies

The training uses the ticked colonies only. Go through the thumbnails of each plate and untick what does not belong: see Looking at colonies. A change to the ticks does not retrain the model; press Train the Classifier again.

How many classes

The SVM needs two or more classes. With one class, it says Switching to one-class SVM for it is only one class defined, and trains a model of that class alone: a colony either belongs or does not. See Training the SVM.

Clusters as classes

Clustering the Test Set replaces the classes with Cluster_1, Cluster_2 and so on, one for each cluster found, which can be trained on like any other classes.