Image Segmentation
Image segmentation, the process of partitioning an image into meaningful regions, aims to accurately delineate objects or areas of interest within a visual scene. Current research heavily emphasizes leveraging foundation models like Segment Anything Model (SAM) and its variants, often incorporating adaptations such as dual-branch architectures or efficient adapters to improve performance on specific domains (e.g., medical imaging, remote sensing) and address limitations like memory consumption. These advancements are significantly impacting diverse fields, from medical diagnosis and industrial inspection to autonomous driving and cultural heritage preservation, by enabling more accurate, efficient, and automated image analysis.
Papers
Average Calibration Error: A Differentiable Loss for Improved Reliability in Image Segmentation
Theodore Barfoot, Luis Garcia-Peraza-Herrera, Ben Glocker, Tom Vercauteren
Toward Robust Canine Cardiac Diagnosis: Deep Prototype Alignment Network-Based Few-Shot Segmentation in Veterinary Medicine
Jun-Young Oh, In-Gyu Lee, Tae-Eui Kam, Ji-Hoon Jeong