The field of 3D scene reconstruction and stylization is moving towards more advanced and realistic methods. Recent developments have focused on improving the fidelity and consistency of 3D scenes generated from single-view images, as well as enabling more controllable and precise style transfer. Notable advancements include the use of Gaussian Splatting and normal bridging to achieve high-fidelity 3D geometry generation and stylization. Additionally, novel frameworks have been proposed to address the challenges of cross-scene object transfer and 3D object arrangement. Noteworthy papers include:

• ABC-GS, which introduces a novel framework for 3D style transfer with controllable matching and style transfer loss functions.
• Hi3DGen, which proposes a framework for generating high-fidelity 3D geometry from images via normal bridging.
• Geometry in Style, which presents a method for identity-preserving mesh stylization via surface normal deformation.
• ExScene, which reconstructs immersive 3D scenes from single-view images using a two-stage pipeline.
• GS-RGBN, which generates high-fidelity 3D objects from single-view images using a hybrid Voxel-Gaussian representation.