Automating Augmentation Selection and Generating Synthetic Microplastics Data to Overcome Small and Imbalanced Data Challenges
GANsemble, a two-module framework, automates augmentation strategy selection and uses the best strategy to train a conditional generative adversarial network (cGAN) to generate high-quality synthetic microplastics data, overcoming challenges posed by small and imbalanced real-world microplastics datasets.