This paper explores the use of a Deep Convolutional Generative Adversarial Network (DCGAN) for data augmentation of grape leaf images to address the challenge of limited training data.
- A DCGAN model was trained on a dataset of 4,062 authentic grape leaf images to generate 16 synthetic samples.
- The quality of generated images was assessed using the Fréchet Inception Distance (FID) score.
- The impact of data augmentation on classification accuracy was evaluated using a pre-trained Inception V3 model.
The paper highlights the results of DCGAN-based data augmentation for grape leaf disease identification, even with a limited number of generated images. The findings contribute to understanding the feasibility and effectiveness of this approach for improving classification accuracy in scenarios with constrained resources.