ATGS: Anchored Temporal Gaussian Splatting for Long Volumetric Video Representation
janusch_patas · x · 2026-09-01
Introduces ATGS (Anchored Temporal Gaussian Splatting), a novel method for long volumetric video representation.
Key Contributions:
- Temporal Anchor-Based Representation: Proposes a representation using time-conditioned anchors and temporal windowing to localize Gaussian primitives, enabling effective modeling of long-sequence dynamic scenes with complex motions.
- Hierarchical Anchor Features: Introduces a formulation that progressively constrains Gaussian generation using global spatial, local spatial, and temporal features, improving spatial stability and temporal consistency.
- Experimental Validation: Conducts extensive experiments on challenging benchmarks, demonstrating consistent outperformance over existing approaches in reconstructing volumetric videos with long durations and complex motions.
More from Research
- AI Research Preference Models: Using Compute to Decide What Experiments to Run — anirudhg9119 · 2026-09-01
- RPM-guided search sets new SOTA on benchmarks, proving selection scales with compute — anirudhg9119 · 2026-09-01
- Inference-only RPM: Frozen LLM Ranks Candidates to Save Compute — anirudhg9119 · 2026-09-01
- From score forecasting to preference ranking: picking which of 15 ideas to run — anirudhg9119 · 2026-09-01
- AI Research Agents Face Bottleneck: Ideas Outpace GPU Budget — anirudhg9119 · 2026-09-01
- Paper: AI Research Preference Models Solve Compute Bottleneck — anirudhg9119 · 2026-09-01