Open Set Semantic Segmentation

Open-set semantic segmentation aims to extend image segmentation beyond pre-defined classes, enabling the identification and classification of unknown objects within a scene. Current research focuses on integrating visual language models, leveraging techniques like mask classification and probabilistic graphical models to handle uncertainty and improve segmentation accuracy for both known and unknown objects. This capability is crucial for advancing robotics, autonomous driving, and other applications requiring robust scene understanding in dynamic and unpredictable environments, particularly where exhaustive training data is unavailable.