QuerySplat: Decoupling Geometry and Appearance in 3DGS Prediction
zhenjun_zhao · x · 2026-08-04
QuerySplat introduces a feed-forward 3D Gaussian Splatting (3DGS) framework driven by geometric priors and explicit appearance decoupling. To address issues like structural redundancy and blurry results in existing methods, the researchers designed a dual-branch query-based decoder. The geometry branch leverages a pretrained Vision Geometric Model for pose-free spatial understanding, while the appearance branch recovers high-frequency details. On the DL3DV benchmark, it achieves state-of-the-art novel view synthesis performance.
More from Multimodal
- MiniMax H3 Test: Flawless Text Rendering and Character Consistency in Video Generation — Exact-Muscle5392 · 2026-08-04
- RTX 5090 Hits 83°C Running MiniMax H3 Locally — Careless-Constant-33 · 2026-08-04
- MiniMax H3 Text-to-Video Successfully Runs on a Single DGX Spark — Scobleizer · 2026-08-04
- Triple MiniMax H3 Speed in ComfyUI with CUDA 13 and INT8 — AI-imagine · 2026-08-04
- Omni Model Generates Complex Animations from a Single Prompt — jnack · 2026-08-04
- Dream Fulfilled: User Transforms into a Mexican Cyclops Using Generative AI — Quirky_Spirit_1951 · 2026-08-04