Synthetic-noise-informed deep learning for excess fraction estimation in circular interferograms

This Zenodo record contains data accompanying the paper "Synthetic-noise-informed deep learning for excess fraction estimation in circular interferograms": source-code.zip is a snapshot of the code repository accompanying the paper; the repository contains code for training the deep learning models, the direct excess fraction method, and synthetic data generation, as well as raw experimental results and visualization scripts ceramic.zip contains interferograms obtained during measurements of the ceramic ball standard described in the paper steel.zip contains interferograms obtained during measurements of the steel standard average_noise.zip is the set of averaged noise images used for creating more realistic synthetic interferograms synthetic_test_data.zip contains synthetic test interferograms with different types of noise (triangular, perlin, black-box, etc.) basic_noise_model.pth is the file with weights of the trained basic noise model perlin_noise_model.pth is the file with weights of the trained Perlin noise model The synthetic data used for model training can be reproduced using the code available in the project's repository and the average noise provided here.