Entity

Why we included this project

Anyone doing image segmentation research or engineering will find this repository useful as a single place that collects several peer-reviewed entity segmentation methods, each with its own README and setup built on Detectron2 and PyTorch. Instead of offering one model, it bundles a family of approaches from the same group's papers: open-world entity segmentation that treats every object as a foreground instance without a fixed class list, a variant for ultra high-resolution images, and a class-agnostic semi-supervised pipeline for detection and segmentation. That breadth helps when comparing segmentation methods, since each sub-project ships its own README, pretrained weights, and evaluation scripts. The code is organized around the academic publications it accompanies, so it also works well as a study aid for understanding how these methods differ. One caveat: the license is non-commercial Creative Commons, so it suits research and experimentation rather than direct commercial deployment.