Installation

This page is only about how to install planktonclass.

If you want the first practical workflow after installation, use Quickstart.

If you want Docker, AI4OS, OSCAR, or the broader project repository, use the companion repository instead:

Option A: Install from PyPI

Standard package install:

pip install planktonclass

Install with GPU support:

pip install "planktonclass[gpu]"

Supported Python versions: 3.10, 3.11, 3.12

What this gives you:

  • the planktonclass command-line tool

  • local training and reporting

  • local DEEPaaS API usage

  • packaged notebook export commands

  • Jupyter notebook runtime dependencies

  • the Python modules used by the package

If you want more detail about GPU support on Windows, Linux, or WSL2, continue to GPU setup.

Option B: Development install

Choose this only if you want to work on the package source itself.

git clone https://github.com/lifewatch/planktonclass
cd planktonclass
python -m venv .venv
.venv\Scripts\activate
pip install -U pip
pip install -e .

After a repository install, you can also start DEEPaaS directly:

$env:planktonclass_CONFIG = (Resolve-Path .\my_project\config.yaml)
$env:DEEPAAS_V2_MODEL = "planktonclass"
deepaas-run --listen-ip 0.0.0.0

Important notes

  • use 127.0.0.1 in the browser; 0.0.0.0 is only the bind address

  • notebook dependencies are included in the default install

  • for training and API usage, you will usually create a project first with planktonclass init my_project

GPU setup

Use:

pip install "planktonclass[gpu]"

After installation, run:

planktonclass doctor

What to look for:

  • TensorFlow runtime: GPU enabled means TensorFlow can use the GPU

  • TensorFlow runtime: GPU unavailable means the current environment is CPU-only

You can also check TensorFlow directly:

python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"

Platform notes:

  • Linux with NVIDIA GPU: primary supported GPU path for training and inference

  • WSL2 on Windows with NVIDIA GPU: recommended Windows-adjacent path for the most future-proof TensorFlow setup

  • Native Windows: GPU support uses DirectML and currently works best with Python 3.10

Python support matrix:

  • CPU on Windows: Python 3.10, 3.11, 3.12

  • CPU on Linux: Python 3.10, 3.11, 3.12

  • GPU on Linux / WSL2: Python 3.10, 3.11, 3.12

  • GPU on native Windows: Python 3.10 only

Native Windows GPU example:

py -3.10 -m venv ..\g310
..\g310\Scripts\python -m pip install --upgrade pip setuptools wheel
..\g310\Scripts\python -m pip install -e ".[gpu]" --no-build-isolation

Or use the helper script from the repository root:

.\scripts\create_gpu_env.ps1

If you hit a Windows long-path installation error, create the environment in a short path such as ..\g310.

Linux or WSL2 GPU example:

python3 -m venv ~/planktonclass-gpu
source ~/planktonclass-gpu/bin/activate
python -m pip install --upgrade pip
pip install "planktonclass[gpu]"

Or use the helper:

./scripts/setup_gpu_linux.sh

Next step

After installation, continue with Quickstart.