F-ANcGAN: An Attention-Enhanced Cycle Consistent Generative Adversarial Architecture for Synthetic Image Generation of Nanoparticles

TL;DR

F-ANcGAN is a novel attention-enhanced GAN architecture that generates realistic SEM images from segmentation maps, addressing data scarcity in nanoparticle analysis and improving downstream segmentation models.

Contribution

We introduce F-ANcGAN, a new GAN model with attention mechanisms and data augmentation for high-quality synthetic nanoparticle images from limited data.

Findings

Achieved a raw FID score of 17.65 on TiO₂ dataset

Reduced FID score to 10.39 with post-processing

Enhanced dataset variety improves segmentation model training

Abstract

Nanomaterial research is becoming a vital area for energy, medicine, and materials science, and accurate analysis of the nanoparticle topology is essential to determine their properties. Unfortunately, the lack of high-quality annotated datasets drastically hinders the creation of strong segmentation models for nanoscale imaging. To alleviate this problem, we introduce F-ANcGAN, an attention-enhanced cycle consistent generative adversarial system that can be trained using a limited number of data samples and generates realistic scanning electron microscopy (SEM) images directly from segmentation maps. Our model uses a Style U-Net generator and a U-Net segmentation network equipped with self-attention to capture structural relationships and applies augmentation methods to increase the variety of the dataset. The architecture reached a raw FID score of 17.65 for TiO$_2$ dataset generation, with a further reduction in FID score to nearly 10.39 by using efficient post-processing techniques. By facilitating scalable high-fidelity synthetic dataset generation, our approach can improve the effectiveness of downstream segmentation task training, overcoming severe data shortage issues in nanoparticle analysis, thus extending its applications to resource-limited fields.