Source Code for DepthCut: Interactive Segmentation with Depth Guidance for Complex Scenarios

This repository provides the source code accompanying the paper “DepthCut: Interactive Segmentation with Depth Guidance for Complex Scenarios.” DepthCut is a depth-guided interactive image segmentation framework that integrates RGB appearance features and depth-based geometric features through an Adaptive Gated Context Fusion module. It aims to improve segmentation in challenging scenarios, including camouflage, underwater scenes, and low-light conditions.

The released code documents the implementation of the proposed method and supports related research and reproduction efforts. This release contains source code and accompanying documentation only; it does not include original image datasets, experimental result data, or pretrained model weights.