RISE-Editing: Rotation-invariant neural point fields with interactive segmentation for fine-grained and efficient editing

Yuze Wang, Junyi Wang, Chen Wang, Yue Qi

Abstract: Highlights•We propose a rotation-invariant neural point field for fine-grained scene editing.•We introduce a multi-view ensemble strategy for 3D segmentation without retraining.•Our framework enables real-time shape editing with simple click-based prompts.•Experiments show superior editing capabilities and photo-realistic novel view synthesis.