Single-Molecule Localization Microscopy (SMLM) 2D Digits 123 and TOL letters and grid dataset processed for machine-learning

The digits and letters dataset was adapted from Huijben, Teun Adrianus Petrus Maria; Heydarian, Hamidreza; Rieger, B. (Bernd); Stallinga, S. (Sjoerd); Jungmann, R. (Ralf) et. al. (2021): Single-Molecule Localization Microscopy (SMLM) 2D Digits 123 and TOL letters datasets. Version 1. 4TU.ResearchData. dataset. https://doi.org/10.4121/14074091.v1 under CC BY-NC 4.0

clusternet_hcf.tar.gz contains the files for ClusterNet-HCF

clusternet_lcf.tar.gz contains the files for ClusterNet-LCF

The salient folders in these are:

config - configuration files 

preprocessed - train and test files in Apache .parquet format - these files could be used to re-train a different network using the pipeline

models - contains the trained model

output - results

processed - train, validation and test files in Pytorch Geometric format

scripts

Note:

the test files are the RESERVED TEST SET FILES

the train and validation set together constitute the data that was used in cross-validation

the clusters contain the handcrafted features.

To reproduce/visualise results follow the instructions at https://github.com/oubino/locpix_points/