Skip to content

Installation

Requirements

Requirement Version Needed for
Python 3.11 or newer Every command
FFmpeg executable A build with the required input codecs and H.264 encoder Event-mode extraction and integration tests
Ultralytics and PyTorch Versions selected by .[detect] Optional object classification

The base package installs PyAV, NumPy, and headless OpenCV. PyAV handles inspection and scanning without a standalone ffmpeg executable. Source-mode extraction also works without it because that mode copies files.

Install from PyPI

Install the released package into an isolated environment.

py -3.12 -m venv .venv
.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip
python -m pip install camreview

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install camreview

CamReview has no standalone binary or container image.

Install from source

Clone the repository and use an editable install when developing or testing an unreleased revision:

git clone https://github.com/willtheorangeguy/camera-review.git
cd camera-review
python -m pip install -e .

Install FFmpeg

Install FFmpeg only on systems that create event-sized clips or run integration tests.

Install a current FFmpeg build using your package manager, then make its bin directory available on PATH.

ffmpeg -version

sudo apt update
sudo apt install -y ffmpeg
ffmpeg -version

Accurate event extraction invokes the libx264 encoder. Confirm that the selected FFmpeg build includes it when accurate extraction fails.

Install object classification

Install the detection dependency group on the machine that runs YOLO:

python -m pip install "camreview[detect]"

The base install doesn't import PyTorch or initialize CUDA. For NVIDIA acceleration, install a PyTorch build compatible with the host driver and CUDA environment, then check it:

python -c "import torch; print(torch.cuda.is_available()); print(torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'CPU')"

--device auto selects cuda:0 when PyTorch reports CUDA availability and otherwise selects CPU. An explicit unavailable CUDA device fails with exit code 5. See Configuration before an offline run because an uncached bare model name may trigger a download.

Install development tools

The development group adds pytest, pytest-cov, Ruff, and mypy:

python -m pip install -e ".[dev]"

Read Testing for test selection and Development guide for the static checks. The base development environment has a documented mypy limitation around optional detection imports.

Verify the installation

Run the console entry point from the activated environment:

camreview --version

CamReview 1.0.1

If the scripts directory isn't on PATH, the module entry point provides the same CLI:

python -m camreview --version

Upgrade

Upgrade a PyPI installation with:

python -m pip install --upgrade camreview

For an editable source installation, pull the desired revision and reinstall with python -m pip install --upgrade -e .. Install .[detect] or .[dev] again when that environment uses the corresponding extra.

Uninstall

Remove the package from the active environment:

python -m pip uninstall camreview

Deleting the virtual environment removes its remaining dependencies. Model weights live outside the environment in the cache described under Configuration, so remove that directory separately when you no longer need the models.