Image Segmentation
Overview
Image segmentation assigns each pixel or voxel to a meaningful region, object, or structure, enabling tasks such as measuring anatomy, mapping land, and understanding scenes. Methods range from thresholding and variational contours to U-Net-style networks, transformers, and foundation models such as SAM, often guided by text, points, boxes, or reference images. Current work also explores learning from sparse labels, synthetic data, multiple modalities, and video, with benchmarks spanning medical scans, natural images, remote sensing, and microscopy. A central challenge is balancing accurate boundaries and topology with reliable performance on new domains, limited annotations, uncertainty, and practical limits on speed and memory.
Latest papers
401 papers24 papers in the last four weeks, up 9% on the four weeks before. 0.1% of all new papers.