Training on multiple files

By default the plugin works on a “layer-level”, i.e. annotations and training are done on a single layer (single image, RGB, stack etc.) However sometimes, one has for example a set of separate images that one wishes to segment in the same way. There might be small differences between images, and in order to capture the full variety of images, it is then necessary to annotate multiple images.

The project tab

By ticking the use project box, the Files tab becomes accessible. This is the place where one can mange the files which are part of the project.

One can drag and drop image to the Files tab. The first time the user is prompted to set a path where to save data. In this location, the plugin saves the list of images added to the project as. By clicking on the added files, they are loaded in the viewer. One can then directly annotate the current image:

Then train a model and do a prediction:

The next file can then be selected and loaded in the viewer and the current model applied to it:

If more annoations are needed, they can be added to the current image. Note that the annotations are automatically saved in the export folder. Therefore, when we select again the first image, the annoations are loaded at the same time:

The entire project can be reloaded later by clicking on Load project.