Class Agnostic

Class-agnostic segmentation aims to identify and delineate objects in images or 3D scenes without prior knowledge of their categories, focusing on separating distinct entities regardless of their class labels. Current research emphasizes developing models that achieve this through bottom-up approaches, leveraging techniques like metric learning and contrastive learning, and employing foundation models such as Segment Anything Model (SAM) for efficient and generalizable segmentation. This capability is crucial for automating tasks like image editing, robotic manipulation (e.g., bin-picking), and analyzing biological assays, offering significant advancements in computer vision and related fields.