Common Vision Studio

English | 简体中文

An open-source edge vision workbench for Raspberry Pi and Linux. Calibrate cameras, run OpenCV/YOLO pipelines, measure world coordinates, control a gimbal, and replay reproducible sessions from one browser.

The media in this repository is generated by the safe synthetic demo. It does not represent measured hardware or model performance.

Why Common Vision Studio?

Most prototypes grow around separate camera loops, one-off calibration scripts, and control processes that cannot reproduce an earlier run. Common Vision Studio provides one shared latest-frame source and a consistent operational boundary:

• camera capture, calibration, source switching, and health diagnostics;
• reusable preprocessing, recognition, geometry, and measurement services;
• navigation, ball tracking, A4 targeting, and a fail-safe gimbal protocol;
• event recording, configuration snapshots, offline replay, and A/B comparison;
• ground-truth evaluation that keeps code regression separate from vision quality;
• a no-hardware demo that never opens a camera or serial port.

Three-minute demo

Python 3.11 or newer is required.

python -m pip install common-vision-studio
common-vision --demo

Open http://127.0.0.1:5000. Demo mode always binds to loopback, uses synthetic frames and an in-process gimbal transport, disables laser authorization, and stores mutable data in an isolated temporary directory.

To work from a clone:

python -m venv .venv
# Linux/macOS: source .venv/bin/activate
# Windows: .venv\Scripts\activate
python -m pip install -e ".[dev]"
common-vision --demo

Feature map

Architecture

flowchart LR
  Sources[Camera / file / stream / synthetic] --> Hub[Shared latest-frame hub]
  Hub --> Vision[Calibration and vision services]
  Vision --> Tasks[Measurement and task pipelines]
  Tasks --> Safety[Expiry, authorization and E-stop]
  Safety --> Gimbal[Gimbal protocol / simulated transport]
  Hub --> Sessions[Record, replay and evaluation]
  Vision --> Sessions

See the architecture guide for ownership, threading, and failure boundaries.

Screenshots

Dashboard	Image preprocessing
Camera calibration	Record and replay

Hardware and platform support

Detailed combinations and reporting requirements are in hardware compatibility.

Installation profiles

pip install common-vision-studio                 # core and safe demo
pip install "common-vision-studio[production]"  # Waitress/Gunicorn
pip install "common-vision-studio[inference]"   # Ultralytics and NCNN

Mutable data defaults to the operating system's application data directory. Use --data-dir PATH or COMMON_VISION_DATA_DIR to select a device data root. Run common-vision-doctor before a real deployment.

Documentation

• Getting started
• Architecture
• Hardware compatibility
• Deployment
• Operations API
• Algorithm validation
• Maintainer release guide
• Roadmap

Contributing and support

Issues and pull requests are welcome. Start with CONTRIBUTING.md and use the structured issue forms so reports include platform and hardware data. Questions belong in GitHub Discussions; security issues must follow SECURITY.md.

The code is licensed under Apache-2.0. Model weights and datasets have separate terms described in MODEL_LICENSES.md.