From Stars to Molecules: AI Guided Device-Agnostic Super-Resolution Imaging

This repository provides code, data, and supplementary material for the paper:
Vašinka et al., From stars to molecules: AI guided device-agnostic super-resolution imaging, Nature Communications 17, 8705 (2026)

This paper is available at Nature Communications.

In addition to this GitHub repository, the DAMN ConvNet and ResNet models, along with the experimental dataset, are stored in the Zenodo repository due to their size.

The repository structure:

Demos

This directory contains code and data for introduction and basic demos on running the DAMN model.

Device_agnostic_training

This directory contains code for training a device-agnostic model using incremental learning with simulated data.

Paper_results

This directory contains all the code and data to reproduce the results presented in the paper.

Technical notes:

All the code was evaluated on Linux with an AMD EPYC 7443P processor and an NVIDIA RTX A5000 GPU using Python 3.11.3. The library versions are included in the requirement .pip files in each individual directory. The computational time denoted in individual files refers to this configuration.
The functionality was also tested on a standalone Linux machine. We did not explicitly test Windows systems, and they probably will not work without a proper Linux emulator due to TensorFlow's GPU support.