This repository provides the official implementation of a dichotomous image segmentation task aimed at highly accurate foreground object extraction from images. It includes IS-Net and U-2-Net based architectures trained for general segmentation use cases, offering inference code for applying the models to custom datasets.