Human-in-the-Loop Atlas-Based 3D Asset Segmentation for Interactive Content Workflows

The article presents a human-in-the-loop pipeline for generating a segmented 2D parameterized atlas from 3D models, enhancing interactive workflows in media, gaming, and XR. The method employs a greedy set cover strategy to select rendered views and utilizes SAM~2 and Label Studio for interactive segmentation, producing unified segmented atlases for tasks like material assignment and style transfer. The evaluation demonstrated effectiveness across diverse geometries, highlighting areas needing manual correction, which is crucial for practitioners aiming to improve segmentation accuracy in complex 3D assets.