Code and Supplementary Materials for: Comparison of Conventional and Generative Methods in Synthetic Data Generation

Repository supporting our paper "Conventional Augmentation is More Effective than ImageGPT and GANs? A Comparison of Synthetic Data Evaluation Methods"

Key files:

• reproduce_results.ipynb: Notebook for reproducing all of the results
• img_transformation.py: Our conventional augmentation function, comprising of transformations from torchvision