Training

To train a new FCN, use the command-line interface (CLI) application binseg train, available on your prompt. To use this CLI, you must define the input dataset that will be used to train the FCN, as well as the type of model that will be trained. You may issue binseg train --help for a help message containing more detailed instructions.

Tip

We strongly advice training with a GPU (using --device="cuda:0"). Depending on the available GPU memory you might have to adjust your batch size (--batch).