Search papers, labs, and topics across Lattice.
This paper introduces ZeroSplat, a novel framework that addresses the limitations of existing Referring 3D Gaussian Splatting (R3DGS) methods by enabling dynamic segmentation of multiple targets in 3D space without the need for extensive feature storage or per-scene optimization. By leveraging multi-view geometric constraints, ZeroSplat effectively translates 2D Vision-Language Model (VLM) priors into a 3D context, achieving intrinsic point-level understanding. Experimental results show that ZeroSplat not only surpasses state-of-the-art methods in both generalized and single-target scenarios but also does so with remarkable computational efficiency.
ZeroSplat achieves dynamic segmentation of multiple targets in 3D without the computational burden of traditional methods, setting a new standard for efficiency and effectiveness in scene understanding.
Recent advancements in 3D Gaussian Splatting (3DGS) have enabled language-guided scene understanding. However, existing Referring 3D Gaussian Splatting (R3DGS) methods are fundamentally restricted to single-target queries. To reflect the ambiguity of real-world instructions, we introduce the Generalized Referring 3D Gaussian Splatting Segmentation (GR3DGS) task, which requires dynamically segmenting an arbitrary number of targets (0, 1, or $N$). To facilitate comprehensive evaluation of this new task, we construct two new benchmarks: GR-LERF and GR-ScanNet. Crucially, existing R3DGS paradigms exhibit fundamental technical bottlenecks that severely limit their performance on the GR3DGS task: they lack intrinsic 3D point-level understanding by operating merely on 2D rendered pixels, and they incur prohibitive computational overhead by requiring per-scene optimization to embed heavy semantic features. To dismantle these bottlenecks, we propose ZeroSplat, a novel training-free and zero-feature framework. ZeroSplat lifts 2D Vision-Language Model (VLM) priors into 3D space through robust multi-view geometric constraints. This strategy enables intrinsic point-level understanding without incurring any additional feature storage. Extensive experiments demonstrate that ZeroSplat significantly outperforms state-of-the-art methods across generalized and single-target scenarios while maintaining exceptional efficiency. Project Page: https://inkmind-ai.github.io/ZeroSplat
