FiftyOne Installation¶

Note

FiftyOne is growing! Sign up for the mailing list to learn about new features as they come out.

Prerequisites¶

You will need a working Python installation. FiftyOne currently requires Python 3.6 - 3.9.

On Linux, we recommended installing Python through your system package manager (APT, YUM, etc.) if it is available. On other platforms, Python can be downloaded from python.org. To verify that a suitable Python version is installed and accessible, run python --version.

We encourage installing FiftyOne in a virtual environment. See setting up a virtual environment for more details.

Installing FiftyOne¶

To install FiftyOne, ensure you have activated any virtual environment that you are using, then run:

pip install fiftyone

This will install FiftyOne and all of its dependencies. Once this has completed, you can verify that FiftyOne is installed in your virtual environment by importing the fiftyone package:

$ python
>>>
>>> import fiftyone as fo
>>>

A successful installation of FiftyOne should result in no output when fiftyone is imported. See this section for install troubleshooting tips.

If you want to work with video datasets, you’ll also need to install FFmpeg:

sudo apt install -y ffmpeg

brew install ffmpeg

You can download a Windows build from here. Unzip it and be sure to add it to your path.

Quickstart¶

Dive right into FiftyOne by opening a Python shell and running the snippet below, which downloads a small dataset and launches the FiftyOne App so you can explore it!

import fiftyone as fo
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")
session = fo.launch_app(dataset)

Note that if you are running this code in a script, you must include session.wait() to block execution until you close the App. See this page for more information.

FiftyOne Desktop App¶

By default, the FiftyOne App will be opened in your web browser when you launch it.

However, we also provide a desktop version of the FiftyOne App that you can install as follows:

pip install fiftyone[desktop]

Note

Commands like launch_app() provide an optional desktop flag that let you control whether to launch the App in your browser or as a desktop App.

You can also set the desktop_app flag of your FiftyOne config to use the desktop App by default.

Troubleshooting¶

If you run into any installation issues, review the suggestions below or check the troubleshooting page for more details.

Note

Most installation issues can be fixed by upgrading some packages and then rerunning the FiftyOne install:

pip install --upgrade pip setuptools wheel
pip install fiftyone

Apple Silicon users:

MongoDB does not yet provide a native build for Apple Silicon, so you must follow these instructions to configure FiftyOne to use a MongoDB instance that you have installed yourself.

Mac users:

You must have the XCode Command Line Tools package installed on your machine. You likely already have it, but if you encounter an error message like error: command 'clang' failed with exit status 1, then you may need to install it via xcode-select --install, or see this page for other options.

Linux users:

The psutil package may require Python headers to be installed on your system. On Debian-based distributions, these are available in the python3-dev package.
If you encounter an error related to MongoDB failing to start, such as Could not find mongod, you may need to install additional packages. See the alternative Linux builds for details.

Windows users:

If you encounter a psutil.NoSuchProcessExists when importing fiftyone, you will need to install the 64-bit Visual Studio 2015 C++ redistributable library. See here for instructions.

Installing extras¶

Various tutorials and guides that we provide on this site require additional packages in order to run. If you encounter a missing package, you will see helpful instructions on what you need to install. Alternatively, you can preemptively install what you’ll need by installing the following additional packages via pip in your virtual environment:

ipython to follow along with interactive examples more easily (note that a system-wide IPython installation will not work in a virtual environment, even if it is accessible)
torch and torchvision for examples requiring PyTorch. The installation process can vary depending on your system, so consult the PyTorch documentation for specific instructions.
tensorflow for examples requiring TensorFlow. The installation process can vary depending on your system, so consult the Tensorflow documentation for specific instructions.
tensorflow-datasets for examples that rely on loading TensorFlow datasets
FFmpeg, in order to work with video datasets in FiftyOne. See this page for installation instructions.

Note

FiftyOne does not strictly require any of these packages, so you can install only what you need. If you run something that requires an additional package, you will see a helpful message telling you what to install.

Upgrading FiftyOne¶

You can upgrade an existing FiftyOne installation by passing the --upgrade option to pip install:

pip install --upgrade fiftyone

If you use the desktop App, you should also run:

pip install fiftyone[desktop]

to ensure that you have the proper version of the desktop App installed for your current FiftyOne version.

Note

New versions of FiftyOne occasionally introduce data model changes that require database migrations after you upgrade. Rest assured, these migrations will be automatically performed on a per-dataset basis whenever you load a dataset for the first time in a newer version of FiftyOne.

Note

If you are working with a custom/shared MongoDB database, you can use database admin privileges to control which clients are allowed to upgrade your FiftyOne deployment.

Downgrading FiftyOne¶

If you need to downgrade to an older version of FiftyOne for any reason, you can do so.

Since new releases occasionally introduce backwards-incompatible changes to the data model, you must use the fiftyone migrate command to perform any necessary downward database migrations before installing the older version of FiftyOne.

Here’s the workflow for downgrading to an older version of FiftyOne:

# The version that you wish to downgrade to
VERSION=0.14.4  # for example

# Migrate the database
fiftyone migrate --all -v $VERSION

# Optional: verify that your datasets were migrated
fiftyone migrate --info

# Now install the older version of `fiftyone`
pip install fiftyone==$VERSION

If you are reading this after encountering an error resulting from downgrading your fiftyone package without first running fiftyone migrate, don’t worry, you simply need to reinstall the newer version of FiftyOne and then follow these instructions.

Note

If you are working with a custom/shared MongoDB database, you can use database admin privileges to control which clients are allowed to downgrade your FiftyOne deployment.

Note

The fiftyone migrate command was introduced in FiftyOne v0.7.3. If you would like to downgrade from a FiftyOne version prior to v0.7.3 (to a yet older version), then you will first need to upgrade to v0.7.3 or later and then follow the instructions above.

Note

To install a FiftyOne version prior to v0.7.0, you must add an --index option to pip install:

pip install --index https://pypi.voxel51.com fiftyone==<version>

Uninstalling FiftyOne¶

FiftyOne and all of its subpackages can be uninstalled with:

pip uninstall fiftyone fiftyone-brain fiftyone-db

If you installed the optional desktop App, you can uninstall it via:

pip uninstall fiftyone-desktop