Folding Drone Dataset: Multi-view Images and Ground Truth Segmentation Masks for 3D Gaussian Splatting

This dataset contains multi-view images and pixel-level Ground Truth (GT) segmentation masks for a real-world "Folding Drone" scene. It is designed to evaluate 3D object segmentation performance within 3D Gaussian Splatting (3DGS) environments, specifically focusing on complex, articulated objects with fine-grained components (e.g., propellers, arms) and severe occlusions. Dataset Contents: Raw multi-view images captured from a real-world scene. Manually annotated Ground Truth (GT) segmentation masks for validation viewpoints. Related Publication: This dataset supports the findings presented in the paper: "HiLOS: Human-in-the-Loop Object Segmentation for 3D Gaussian Splatting via Finger-based VR Interface" (submitted to Advanced Intelligent Systems, 2025).