Tile-Wise vs. Image-Wise: Random-Tile Loss and Training Paradigm for Gaussian Splatting
Abstract
3D Gaussian Splatting (3DGS) has drawn significant attention for its advantages in rendering speed and quality. Most existing methods still rely on the image-wise loss and training paradigm because of its intuitive nature in the Splatting algorithm. However, image-wise loss lacks multi-view constraints, which are generally essential for optimizing 3D appearance and geometry. To address this, we propose RT-Loss along with a tile-based training paradigm, which uses randomly sampled tiles to integrate multi-view appearance and structural constraints in 3DGS. Additionally, we introduce an tile-based adaptive densification control strategy tailored for our training paradigm. Extensive experiments show that our approach consistently improves performance metrics while maintaining efficiency across various benchmark datasets.
Cite
Text
Zhang et al. "Tile-Wise vs. Image-Wise: Random-Tile Loss and Training Paradigm for Gaussian Splatting." International Conference on Computer Vision, 2025.Markdown
[Zhang et al. "Tile-Wise vs. Image-Wise: Random-Tile Loss and Training Paradigm for Gaussian Splatting." International Conference on Computer Vision, 2025.](https://mlanthology.org/iccv/2025/zhang2025iccv-tilewise/)BibTeX
@inproceedings{zhang2025iccv-tilewise,
title = {{Tile-Wise vs. Image-Wise: Random-Tile Loss and Training Paradigm for Gaussian Splatting}},
author = {Zhang, Xiaoyu and Pan, Weihong and Xiang, Xiaojun and Zhai, Hongjia and Zhou, Liyang and Jiang, Hanqing and Zhang, Guofeng},
booktitle = {International Conference on Computer Vision},
year = {2025},
pages = {26923-26932},
url = {https://mlanthology.org/iccv/2025/zhang2025iccv-tilewise/}
}