openretina
Welcome!
openretina is an open-source neural network toolkit for modeling retinal responses to visual stimuli. It provides pre-trained models for inference and interpretability, as well as components for training new retina models with your own data.
-
Check out the main package repository on GitHub
-
openretinahas a biorxiv pre-print! -
Default, most flexible option. Comes with config files and notebooks.
git clone git@github.com:open-retina/open-retina.git cd open-retina pip install -e .Simplest installation for using pre-trained models
pip install openretina -
Browse and download retinal datasets from our HuggingFace repository.
Quick Start
import torch
from openretina.models import load_core_readout_from_remote
# Load a pre-trained model
model = load_core_readout_from_remote("hoefling_2024_base_low_res", "cpu")
# Run a forward pass with random input
responses = model.forward(torch.rand(model.stimulus_shape(time_steps=50)))
print(f"Model predicted responses shape: {responses.shape}")
Participating Labs
-
Euler Lab
University of Tübingen, Center for Integrative Neuroscience
-
Bethge Lab
University of Tübingen, AI Center
-
Marre Lab
Institut de la Vision, Paris
-
Ecker Lab
University of Göttingen
-
Gollisch Lab
University Medical Center Göttingen, Department of Ophthalmology
- Lead maintainers
- Federico D'Agostino — University of Tübingen, AI Center
Thomas Zenkel — University of Tübingen, Center for Integrative Neuroscience
Larissa Höfling — University of Tübingen, AI Center - Contributors
- Baptiste Lorenzi — Institut de la Vision, Paris
Michaela Vystrčilová — University of Göttingen
Dominic Gonschorek — University of Tübingen, Center for Integrative Neuroscience
Samuel Suhai — University of Tübingen, Center for Integrative Neuroscience
Shashwat Sridhar — University Medical Center Göttingen
Samuele Virgili — Institut de la Vision, Paris