iiinova — where your digital self keeps evolving

Thomas Zenkel

@thomaszen.bsky.social

Neuroscientist in Tübingen

Thomas Zenkel @thomaszen.bsky.social · 23/07/2026

With @fededagos.bsky.social @dgonschorek.bsky.social @svirgili.bsky.social @aecker.bsky.social @oliviermarre.bsky.social @lhoefling.bsky.social @teuler.bsky.social @bethgelab.bsky.social Baptiste Lorenzi, Michaela Vystrčilová, Samuel Suhai, and hopefully many more contributors to come.

000

Thomas Zenkel @thomaszen.bsky.social · 23/07/2026

We envision openretina as a collaborative effort. Contribute datasets or models, use pretrained checkpoints as baselines, and test whether improvements generalise across datasets and species.

100

Thomas Zenkel @thomaszen.bsky.social · 23/07/2026

openretina additionally includes tools for in silico analysis of trained models like stimulus optimisation, weight visualisation, response-tuning maps, and gradient-field analyses.

100

Thomas Zenkel @thomaszen.bsky.social · 23/07/2026

openretina provides a reproducible training pipeline and unified evaluation metrics for benchmarking models within and across datasets. The current benchmarks also show that substantial explainable variance remains uncaptured by existing models.

100

Thomas Zenkel @thomaszen.bsky.social · 23/07/2026

The models build on the widely used Core + Readout architecture: a shared Core extracts visual features, while neuron-specific Readouts map those features to predicted responses for individual retinal ganglion cells.

100

Thomas Zenkel @thomaszen.bsky.social · 23/07/2026

So far, we integrate five public datasets, spanning species, recording modalities, and stimulus types. Each includes curated preprocessing, standardised data loaders, and pretrained model checkpoints. Data and models: huggingface.co/datasets/ope...

huggingface.co

open-retina/open-retina · Datasets at Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

110

Thomas Zenkel @thomaszen.bsky.social · 23/07/2026

openretina brings neural network models, that are trained to predict the activity of retinal ganglion cells in response to visual stimuli, into a shared framework and bundles datasets, pretrained models, and tools for in silico analysis.

100

Thomas Zenkel @thomaszen.bsky.social · 23/07/2026

📣 Our paper introducing openretina, a collaborative framework for retina modelling across datasets and species, is now published as an eLife Reviewed Preprint. Paper: elifesciences.org/reviewed-pre... Code: github.com/open-retina/... Docs: open-retina.org

12911

Reposted by Thomas Zenkel

Federico D’Agostino @fededagos.bsky.social · 14/03/2025

🚨 New paper alert! 🚨 We’ve just launched openretina, an open-source framework for collaborative retina modeling across datasets and species. A 🧵👇 (1/9)

13820