让小显存设备也能训练超大3D高斯点云，靠分块更新和预合成。
ABCD: Alpha-Composited Block Coordinate Descent: Constant-VRAM Training for Large Radiance Fields
- 分块更新参数，每次只激活一块，其余冻结并预合成图像。
- 显存占用恒定，不随场景大小增长，实测峰值显存仅1.2GB。
- 适合资源有限但需重建复杂场景的研究者或开发者。
我们提出ABCD（Alpha-Composited Block Coordinate Descent），一种用于alpha合成辐射场的外部内存训练框架，此处以3D高斯点云（3DGS）为例。该方法将训练重构为基于空间分区的块坐标下降：每次仅激活一个参数块，其余冻结。利用alpha混合的结合律，这些非活跃区域可预先渲染并合并为前景与背景的RGBA图像。因此，在固定分区大小与图像分辨率下，峰值显存需求对总场景范围呈常数级，而非随全场景规模增长。这使得显存受限的GPU也能训练原本无法容纳的大型场景。实验表明，本方法在3DGS重建质量上几乎无损，PSNR下降不足5%；而移除合成步骤后则退化约40%。代码见https://github.com/shiukaheng/abcd。
原文摘要 · Abstract (English)
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
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。