Software

Software setup, configuration, and model-zoo.

The MEYELens library is designed to support acquisition, pupil detection, gaze-related workflows, and export of reproducible outputs.

Installation Selector

Select your MEYELens configuration and copy the generated install command.

Tip: Tutorials are available in the Resources page .

Complete Installation Instructions

The library is written in Python. We recently moved from TensorFlow to PyTorch for performance reasons, especially to improve inference speed and flexibility across different hardware configurations. We recommend installing MEYELens in a dedicated Python environment.

Recommended install

Regardless of the MEYELens version you want to install, we recommend creating a dedicated Python environment first. This keeps the installation clean and helps avoid conflicts with other packages. After creating and activating the environment, you can install the version that best matches your setup.

Command

conda create -n meyelens python=3.11 -y
conda activate meyelens

CPU version

The easiest way to install MEYELens is to use the CPU version. This installation includes the standard PyTorch dependency set and does not require any manual GPU or CUDA configuration. It is the recommended option if you do not have a supported GPU, or if you simply want a straightforward setup that works on most computers. The CPU version is complete and can run the full MEYELens workflow, although predictions may run at a lower frame rate than with GPU acceleration.

Command

pip install "meyelens[pt]"

PyTorch GPU support (optional)

If you want GPU acceleration, configure PyTorch first by following the official installation guide for your platform. After verifying that PyTorch can access the GPU, install MEYELens:

Command

pip install meyelens

Warning: GPU acceleration depends on your operating system, CUDA/cuDNN compatibility, and PyTorch version. Follow the official Pytorch installation guide for your platform before installing MEYELens.

Tip: This install MEYELens alongside your manually configured PyTorch library.

Tip: After installing PyTorch with GPU support, verify that Python can access the GPU with the command below.

python -c "import torch; print('CUDA:', torch.cuda.is_available()); print('GPU:', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'None')"

Expected result: CUDA: True, followed by the name of your GPU. If the result is CUDA: False, PyTorch is installed but it cannot access the GPU, so MEYELens will run on CPU.

Headless package

If you need only the core Python package without the PyQt GUI, install the headless version. This option is particularly useful on smaller or dedicated devices, such as NVIDIA Jetson boards , embedded systems, remote acquisition computers, or other setups where a graphical interface is not required.

Command

pip install meyelens-headless

To install the headless package with PyTorch dependencies (for CPU):

Command

pip install "meyelens-headless[pt]"

Warning: Do not install meyelens and "meyelens-headless" into the same environment. Use separate environments if you need to test both distributions.

Documentation: API documentation is available here .

Legacy TensorFlow version

The original MEYE TensorFlow model is no longer actively maintained, because current MEYELens development is moving toward the newer PyTorch-based pipeline. However, the TensorFlow model remains available for reproducibility and for users who rely on the legacy MEYE workflow. You can try it through the pupillometry.it web app, download the pretrained model here, or consult the GitHub project and the original paper for implementation details.

The same legacy model is also listed in the MEYELens Model Zoo, together with its main specifications and usage notes.

Model Zoo

Pretrained models and reusable model assets for MEYELens analysis workflows.