# TimeFormer: Capturing Temporal Relationships of Deformable 3D Gaussians for Robust Reconstruction

arXiv: https://arxiv.org/abs/2411.11941 · HTML: https://mlsanity.com/paper/2411.11941/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Builds on 62 of its 75 references that are in the map.
- First posted on arXiv 18 Nov 2024.

## Builds on (most cited first, 30 of 62)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [3D Gaussian Splatting for Real-Time Radiance Field Rendering](https://mlsanity.com/paper/2308.04079/) (2023)
- [Instant Neural Graphics Primitives with a Multiresolution Hash Encoding](https://mlsanity.com/paper/2201.05989/) (2022)
- [RAFT: Recurrent All-Pairs Field Transforms for Optical Flow](https://mlsanity.com/paper/2003.12039/) (2020)
- [Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields](https://mlsanity.com/paper/2103.13415/) (2021)
- [Plenoxels: Radiance Fields without Neural Networks](https://mlsanity.com/paper/2112.05131/) (2021)
- [D-NeRF: Neural Radiance Fields for Dynamic Scenes](https://mlsanity.com/paper/2011.13961/) (2020)
- [TensoRF: Tensorial Radiance Fields](https://mlsanity.com/paper/2203.09517/) (2022)
- [Nerfies: Deformable Neural Radiance Fields](https://mlsanity.com/paper/2011.12948/) (2020)
- [4D Gaussian Splatting for Real-Time Dynamic Scene Rendering](https://mlsanity.com/paper/2310.08528/) (2023)
- [Direct Voxel Grid Optimization: Super-fast Convergence for Radiance Fields Reconstruction](https://mlsanity.com/paper/2111.11215/) (2021)
- [PlenOctrees for Real-time Rendering of Neural Radiance Fields](https://mlsanity.com/paper/2103.14024/) (2021)
- [Deformable 3D Gaussians for High-Fidelity Monocular Dynamic Scene Reconstruction](https://mlsanity.com/paper/2309.13101/) (2023)
- [Neural Scene Flow Fields for Space-Time View Synthesis of Dynamic Scenes](https://mlsanity.com/paper/2011.13084/) (2020)
- [K-Planes: Explicit Radiance Fields in Space, Time, and Appearance](https://mlsanity.com/paper/2301.10241/) (2023)
- [Dynamic 3D Gaussians: Tracking by Persistent Dynamic View Synthesis](https://mlsanity.com/paper/2308.09713/) (2023)
- [HexPlane: A Fast Representation for Dynamic Scenes](https://mlsanity.com/paper/2301.09632/) (2023)
- [Real-time Photorealistic Dynamic Scene Representation and Rendering with 4D Gaussian Splatting](https://mlsanity.com/paper/2310.10642/) (2023)
- [Non-Rigid Neural Radiance Fields: Reconstruction and Novel View Synthesis of a Dynamic Scene From Monocular Video](https://mlsanity.com/paper/2012.12247/) (2020)
- [Dynamic View Synthesis from Dynamic Monocular Video](https://mlsanity.com/paper/2105.06468/) (2021)
- [Space-time Neural Irradiance Fields for Free-Viewpoint Video](https://mlsanity.com/paper/2011.12950/) (2020)
- [SC-GS: Sparse-Controlled Gaussian Splatting for Editable Dynamic Scenes](https://mlsanity.com/paper/2312.14937/) (2023)
- [Fast Dynamic Radiance Fields with Time-Aware Neural Voxels](https://mlsanity.com/paper/2205.15285/) (2022)
- [Spacetime Gaussian Feature Splatting for Real-Time Dynamic View Synthesis](https://mlsanity.com/paper/2312.16812/) (2023)
- [Neural Radiance Flow for 4D View Synthesis and Video Processing](https://mlsanity.com/paper/2012.09790/) (2020)
- [Tensor4D : Efficient Neural 4D Decomposition for High-fidelity Dynamic Reconstruction and Rendering](https://mlsanity.com/paper/2211.11610/) (2022)
- [Gaussian-Flow: 4D Reconstruction with Dynamic 3D Gaussian Particle](https://mlsanity.com/paper/2312.03431/) (2023)
- [Robust Dynamic Radiance Fields](https://mlsanity.com/paper/2301.02239/) (2023)

## Built on by (most cited first, 9 of 9)

- [Advances in 4D Representation: Geometry, Motion, and Interaction](https://mlsanity.com/paper/2510.19255/) (2025)
- [Self-supervised Normality Learning and Divergence Vector-guided Model Merging for Zero-shot Congenital Heart Disease Detection in Fetal Ultrasound Videos](https://mlsanity.com/paper/2503.07799/) (2025)
- [The Semantic Lifecycle in Embodied AI: Acquisition, Representation and Storage via Foundation Models](https://mlsanity.com/paper/2601.08876/) (2026)
- [Mango-GS: Enhancing Spatio-Temporal Consistency in Dynamic Scenes Reconstruction using Multi-Frame Node-Guided 4D Gaussian Splatting](https://mlsanity.com/paper/2603.11543/) (2026)
- [On the Design of Mixture-of-Experts for Dynamic Gaussian Splatting](https://mlsanity.com/paper/2607.08250/) (2026)
- [ReFlow: Self-correction Motion Learning for Dynamic Scene Reconstruction](https://mlsanity.com/paper/2604.01561/) (2026)
- [FreeTimeGS++: Secrets of Dynamic Gaussian Splatting and Their Principles](https://mlsanity.com/paper/2605.03337/) (2026)
- [DLGStream: Dynamic Language-embedded Guassian Splatting for Open-vocabulary Enabled Free-viewpoint Video Streaming](https://mlsanity.com/paper/2606.28840/) (2026)
- [MVFusion-GS: Motion-Variance Guided Temporal Attention for High-Quality Dynamic Gaussian Splatting](https://mlsanity.com/paper/2607.01578/) (2026)

## Abstract (arXiv)

Dynamic scene reconstruction is a long-term challenge in 3D vision. Recent methods extend 3D Gaussian Splatting to dynamic scenes via additional deformation fields and apply explicit constraints like motion flow to guide the deformation. However, they learn motion changes from individual timestamps independently, making it challenging to reconstruct complex scenes, particularly when dealing with violent movement, extreme-shaped geometries, or reflective surfaces. To address the above issue, we design a plug-and-play module called TimeFormer to enable existing deformable 3D Gaussians reconstruction methods with the ability to implicitly model motion patterns from a learning perspective. Specifically, TimeFormer includes a Cross-Temporal Transformer Encoder, which adaptively learns the temporal relationships of deformable 3D Gaussians. Furthermore, we propose a two-stream optimization strategy that transfers the motion knowledge learned from TimeFormer to the base stream during the training phase. This allows us to remove TimeFormer during inference, thereby preserving the original rendering speed. Extensive experiments in the multi-view and monocular dynamic scenes validate qualitative and quantitative improvement brought by TimeFormer. Project Page: https://patrickddj.github.io/TimeFormer/