NeuMesh++: Towards Versatile and Efficient Volumetric Editing with Disentangled Neural Mesh-based Implicit Field
Researchers have developed NeuMesh++, a novel mesh-based representation for neural implicit rendering that enhances editing capabilities for 3D scenes. This method disentangles geometry, texture, and semantic information onto mesh vertices, enabling versatile operations such as mesh-guided geometry editing, texture swapping, filling, painting, and semantic-guided editing. The system also incorporates techniques like local space parameterization and learnable vertex colors to improve rendering quality and texture editing fidelity, with experiments demonstrating its effectiveness on both real and synthetic datasets. AI
IMPACT Enhances 3D content creation and editing workflows with more intuitive and versatile tools.
- arXiv
- Neural radiance field
- geometry
- Novel View Synthesis
- 3D Scene Reconstruction Using Omnidirectional Vision and LiDAR: A Hybrid Approach
- NeuMesh++
- neural implicit rendering
- mesh-based representation
- surface roughness
- mesh vertices
- mesh-guided geometry editing
- texture swapping
- filling and painting operations
- semantic-guided editing