Trained ESRGAN models for single-frame super-resolution microscopy (STORM-like image generation).

This dataset contains the model weights (PyTorch) of the deep learning models (ESRGAN) described in the paper: Generative super-resolution AI accelerates nanoscale analysis of cells (2025). These models, specifically A, B1, and B2, were developed to generate STORM-like super-resolution images from single-frame widefield microscopy inputs. The models were trained using a 4× upscale factor on a dataset of microtubules from human cell lines, collected at the CISUP Laboratory of the University of Pisa. Paricularly, they were trained to generate STORM-like super-resolution images starting from widefield microscopy inputs.