Boosting Single-View Scene Completion via Knowledge Distillation from Multi-View Reconstruction
The core message of this work is to boost the performance of single-view scene completion by leveraging knowledge distillation from a multi-view scene reconstruction model trained in a fully self-supervised manner.
Leveraging Semantic and Spatial Context for Improved Single-View 3D Scene Reconstruction
By incorporating fine-grained semantic information and reasoning about 3D spatial context, our method KYN achieves state-of-the-art performance in single-view scene and object reconstruction, especially in occluded regions.
Enhanced Indoor 3D Scene Reconstruction with Occluded Surface Completion
A novel indoor 3D reconstruction method that completes occluded surfaces in addition to reconstructing visible surfaces, using a hierarchical octree representation and a dual-decoder architecture.
Mirror-3DGS: Accurately Rendering Mirror Reflections in 3D Gaussian Splatting
Mirror-3DGS innovatively incorporates mirror imaging principles into 3D Gaussian Splatting, enabling accurate and realistic rendering of mirror reflections in real-time.
© 2024 by Linnk AI