Fractal-NST: Fractal-Based Neural Style Transfer Image Augmentation

This repository provides sample data generated using the proposed Fractal-based Neural Style Transfer (Fractal-NST) framework, which is designed for high-quality and unbiased image augmentation. The shared data are intended to support transparency, reproducibility, and qualitative evaluation of the methods presented in the associated research paper.

The repository is organized into two primary directories:

The Fractal Images directory contains fractal-based style images generated using multiple fractal generation techniques. Fractals are mathematically defined, non-semantic patterns that exhibit self-similarity and rich textural characteristics. For each fractal generation technique, 200 images are provided and stored in separate subfolders to ensure clarity and ease of use. These images serve as style sources for neural style transfer and are domain-independent, making them suitable for augmentation across different datasets without introducing semantic or class bias.

The Augmented Images directory contains 100 randomly selected augmented images produced using the proposed Fractal-NST method. These images demonstrate the effect of transferring fractal-based styles onto original content images while preserving essential structural and semantic information. The samples highlight the diversity, texture richness, and content fidelity achieved by the proposed approach and are representative of the augmented datasets used during experimental evaluation.

All images are provided in standard image formats and maintain consistent spatial resolution with the original datasets used in the study. The repository does not include original datasets or sensitive data; only generated fractal images and representative augmented samples are shared. The data can be used for visual inspection, methodological validation, and comparative analysis of neural style transfer-based augmentation techniques.

The Fractal-NST implementation code is available at: https://github.com/kanch-git/Fractal-NST/

References:

https://doi.org/10.17632/tywbtsjrjv.1

https://doi.org/10.1016/j.dib.2023.109306