Few-Shot Hyperspectral Aphid Detection via FastGAN Synthetic Data Generation, Transformer-Based Classification and Explainable AI

The study introduces a data-efficient generative adversarial network (FastGAN) to augment a hyperspectral dataset for early detection of aphid infestations in crops, generating 10,000 synthetic images that maintain the structural and spectral integrity of real samples. The augmented dataset was utilized to train several classification architectures, including VGG16, ResNet-50, EfficientNet, and Vision Transformer (ViT), with the ViT model achieving the highest accuracy and F1-scores. This research highlights the effectiveness of FastGAN in enhancing classification robustness for hyperspectral imaging, particularly emphasizing the superior performance of transformer-based models in distinguishing between healthy and infested plant leaves.