MLP Splatting: Object-Centric Neural Fields
Researchers have introduced MLP-Splatting, a novel method for creating object-centric 3D scene representations. This technique utilizes compact Multi-Layer Perceptron (MLP) primitives, each with localized spatial support, to model radiance and opacity. Unlike previous methods, MLP-Splatting allows for easier scene decomposition into object-level primitives without requiring explicit segmentation masks, enabling interactive editing. The method also boasts significantly reduced memory usage and faster rendering times compared to existing state-of-the-art approaches. AI
IMPACT Enables more intuitive object-level manipulation and editing of 3D scenes, potentially impacting AR/VR and content creation.