Segmentation models are typically constrained by the categories defined during training. To
address this, researchers have explored two independent approaches: adapting Vision …