ABCD：Alpha合成块坐标下降法：针对大辐射场的恒定显存训练
ABCD: Alpha-Composited Block Coordinate Descent: Constant-VRAM Training for Large Radiance Fields
浏览论文内容
中文总结 AI 辅助
该研究提出ABCD外核训练框架，针对3DGS实现，通过块坐标下降与Alpha混合特性将峰值显存复杂度降为O(1)，在PSNR下降不足5%的前提下，让显存有限的GPU可训练大辐射场
中文摘要 AI 辅助
我们提出了ABCD（Alpha合成块坐标下降法，Alpha-Composited Block Coordinate Descent），这是一种用于Alpha合成辐射场的 out-of-core（外核）训练框架，本文中针对3D Gaussian Splatting（3DGS）实现了该框架。我们的方法将训练重新表述为对空间划分的块坐标下降：每次仅激活一个参数块，其余所有块保持冻结。通过利用Alpha混合的结合性，这些非活动区域可被预渲染并折叠为前景和背景RGBA图像。因此，对于固定的划分大小和图像分辨率，峰值显存（VRAM）复杂度为O(1)，与总场景范围无关，而非随整个场景大小增长。这使得显存有限的GPU能够训练原本无法放入核心内存的场景。实验中，我们的方法紧密保留了3DGS的重建质量，PSNR（峰值信噪比）下降不到5%；而去除合成的ABCD则会出现约40%的下降。我们的代码可在此处获取：this https URL
英文摘要
We present ABCD (Alpha-Composited Block Coordinate Descent), an out-of-core training framework for alpha-composited radiance fields, instantiated here for 3D Gaussian Splatting. Our method reformulates training as block coordinate descent over spatial partitions: only one block of parameters is active at a time, while all others are frozen. By exploiting the associativity of alpha blending, these inactive regions can be pre-rendered and collapsed into foreground and background RGBA images. As a result, for fixed partition size and image resolution, peak VRAM becomes O(1) with respect to total scene extent, rather than growing with full scene size. This enables GPUs with limited memory to train scenes that would otherwise not fit in core. In experiments, our method closely preserves the reconstruction quality of 3DGS, with less than 5% PSNR degradation, while ABCD with compositing ablated suffers roughly 40% degradation. Our code can be found at https://github.com/shiukaheng/abcd
发表机构
- The University of Edinburgh（爱丁堡大学）
机构由 AI 辅助整理，请以论文原文为准。
