INSAIT, in collaboration with leading international research institutions, is proud to announce the release of SceneSplat—the first system for instant 3D object discovery from text within photorealistic scenes. With SceneSplat, users can search for any phrase, not just a predefined set of categories, and instantly identify relevant objects directly in 3D space.
Built on top of the revolutionary 3D Gaussian Splatting technology, SceneSplat sets a new benchmark in 3D scene understanding. Unlike traditional systems that require slow, scene-specific optimization—often taking minutes per scene—SceneSplat processes entire scenes in seconds, thanks to a single forward pass through a neural network.
SceneSplat is also the first large-scale and generalizable method for indoor 3D scene understanding that natively operates on 3D Gaussian Splatting. It assigns open-vocabulary language features to each 3D Gaussian element, enabling rich, flexible, and intuitive language-based interaction within complex 3D environments.
This innovative work is the result of a collaboration between INSAIT, the University of Amsterdam, ETH Zurich (Computer Vision Lab), Nanjing University of Aeronautics and Astronautics, Università di Pisa, and Università di Trento.
Congratulations to the authors: Yue Li, Qi Ma, Runyi Yang, Huapeng Li, Mengjiao Ma, Bin Ren, Nikola Popovic, Nicu Sebe, Ender Konukoglu, Theo Gevers, Luc Van Gool, Martin R. Oswald, and Danda Pani Paudel.