Reproduce “Cross3DGS: Unified Vulkan-Based 3D Gaussian Splatting Training from Workstations to Mobile Devices” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hiroyuki Sato, Yusuke Takimoto
Cross3DGS trains 3D Gaussian Splatting scenes on heterogeneous devices by retaining one optimizer and selecting a device-appropriate GPU execution schedule. It filters 256 candidate schedules using device limits, workload, and driver rules, then applies a fixed preference order without timing them on-device. It completed 30,000 updates on all eight tested configurations; quality was similar on two scenes, not universally.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.