界面已本地化;技能介绍与说明文档按所选语言由 AI 自动翻译,代码、命令、路径和专业术语保留原文,以原作者文档为准。
探索技能
效率与协作 · SKILL.md

live-demo

live-demo

【效率与协作】Run and troubleshoot MonoGS RealSense live demos and the Open3D/OpenGL GUI.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准

技能说明

GitHub

name: live-demo description: "Run and troubleshoot MonoGS RealSense live demos and the Open3D/OpenGL GUI." disable-model-invocation: true metadata: disco-role: operating license: NOASSERTION

Live Demo

Use this sub-skill when the user asks about live MonoGS capture, Intel RealSense cameras, USB/device access, GUI behavior, Open3D/GLFW/OpenGL display errors, or interactive visualization controls.

Route away

  • Base CUDA/PyTorch/submodule installation belongs to environment-setup.
  • Offline TUM/Replica/EuRoC data and configs belong to data-and-configs.
  • Offline SLAM commands belong to offline-slam.
  • Saved-result metrics belong to evaluation-and-results.

Fast path

  1. Verify the base MonoGS environment and CUDA extensions first.
  2. Install the optional RealSense dependency only when the user will run live camera workflows:
    pip install pyrealsense2
    
  3. Check GUI and camera prerequisites without opening a window:
    python scripts/check_live_demo_prereqs.py --require-cuda
    
    Add --probe-camera only with an attached RealSense device and user approval.
  4. Choose a live config:
    • monocular RealSense: python slam.py --config configs/live/realsense.yaml
    • RGB-D RealSense: python slam.py --config configs/live/realsense_rgbd.yaml
  5. Use USB-3, avoid aggressive motion during initialization, and expect GUI to be forced on for Dataset.type: realsense.

Bundled references

我的收藏