Installation

FlexCloud ships both as a PyPI wheel (recommended for most users) and as a buildable source tree with a pre-built Docker image.

Standalone pip-package

The simplest path:

pip install flexcloud

Use a virtual environment

Install into a virtual environment (or pipx) instead of your system Python to avoid dependency clashes and keep the install fully removable:

python -m venv .venv
source .venv/bin/activate
pip install flexcloud

Rerun viewer required for visualization

The visualization output is streamed to the Rerun viewer, which is not pulled in by the wheel — you have to install it yourself on the host:

pip install rerun-sdk

Without it the algorithms still run, but you will not see any 3D output. The Docker image bundles its own Rerun viewer (see below), so this step is only needed for the standalone pip install.

Verifying the installation

flexcloud-keyframe-interpolation --help
flexcloud-georeferencing --help

Both commands should print their usage banner.

Docker

A containerised build is provided for users who prefer not to install system dependencies.

# build locally
./docker/build_docker.sh

# or pull a prebuilt image
docker pull ghcr.io/tumftm/flexcloud:latest

# run, mounting your data directory
./docker/run_docker.sh /your/local/directory/data

The container ships with a Rerun viewer instance; to use it, you need to activate the virtual environment in the container that it is installed in:

source /ros_ws/rerun/bin/activate

If you encounter graphics-driver issues you can launch a Rerun viewer on the host instead — see Test Data & Visualization.