Search papers, labs, and topics across Lattice.
FLUX3D addresses the limitations of existing sparse voxel representations in 3D Gaussian Splatting (3DGS) by enhancing both representation learning and cross-modal alignment. The framework introduces Diffusion-Aligned Structured Latents (DA-SLAT) and a novel sparse-structure-aware diffusion architecture, which collectively improve the fidelity of 3D reconstructions. Benchmark experiments reveal that FLUX3D significantly surpasses state-of-the-art methods in generating high-quality 3DGS assets, particularly in preserving high-frequency visual details.
FLUX3D achieves unprecedented fidelity in 3D Gaussian generation by effectively aligning 2D image tokens with sparse 3D voxel latents.
Sparse voxel representation has emerged as a scalable foundation for image-to-3D Gaussian Splatting (3DGS) generation, yet current methods struggle to preserve high-frequency visual details of input images due to two structural bottlenecks. First, they adopt discriminative 2D features optimized for semantic abstraction to construct sparse voxel latents, which suppress reconstructive cues and induce a representation bottleneck. Second, in the generation stage, standard diffusion transformers lack effective mechanisms to align dense 2D image tokens with sparse 3D voxel latents, resulting in a cross-modal correspondence bottleneck. To address these issues, we propose FLUX3D, a scalable image-to-3DGS framework that boosts both representation learning and cross-modal alignment during generation. We first revisit 2D feature selection for sparse-voxel-based 3D representation learning, propose Diffusion-Aligned Structured Latents (DA-SLAT) and couple it with a decoder-only architecture to improve 3DGS reconstruction fidelity. We also design a sparse-structure-aware diffusion framework, which integrates the Sparse-structure Multimodal Diffusion Transformer (SMDiT) and Modal-Aware Rotary Positional Embedding (MARoPE) to achieve geometry-agnostic 2D-3D alignment. Extensive benchmark experiments demonstrate that FLUX3D yields substantial improvements in appearance fidelity and significantly outperforms all state-of-the-art (SOTA) methods in generating high-quality 3DGS assets.
