Training-free zero-shot 3D symmetry detection with visual features back-projected to geometry
This work addresses symmetry detection in 3D objects for computer vision and graphics applications, offering a novel zero-shot approach that leverages pre-trained models without additional training.
The paper tackled 3D symmetry detection by proposing a training-free method that uses visual features from foundation models like DINOv2, back-projecting them to geometry to identify reflection-symmetry planes, and it outperformed traditional and learning-based approaches on a ShapeNet subset.
We present a simple yet effective training-free approach for zero-shot 3D symmetry detection that leverages visual features from foundation vision models such as DINOv2. Our method extracts features from rendered views of 3D objects and backprojects them onto the original geometry. We demonstrate the symmetric invariance of these features and use them to identify reflection-symmetry planes through a proposed algorithm. Experiments on a subset of ShapeNet demonstrate that our approach outperforms both traditional geometric methods and learning-based approaches without requiring any training data. Our work demonstrates how foundation vision models can help in solving complex 3D geometric problems such as symmetry detection.