Exploring Image Augmentation Techniques for Enhanced Deep Learning-based Plant Disease Identification

Agriculture stands as the main source of essential vegetation crucial for supporting the growing global population. The prevalence of different plant diseases emphasizes how important it is to diagnose crop diseases as soon as possible. Nevertheless, the traditional approaches to disease identification turn out to be costly, labor-intensive, and dependent on a thorough knowledge of plant diseases. Machine learning (ML) and deep learning (DL) techniques have been the focus of recent studies in an effort to overcome these obstacles for automated crop disease identification. DL algorithms excel at diagnosing diseases, although overfitting can occur due to a lack of labeled data, reducing their usefulness on unlabeled data. To tackle this problem, data augmentation techniques are utilized to increase the dataset size without requiring more field image data. This study examines image augmentation methods for identifying plant diseases, with a focus on deep learning-based techniques such as neural style transfer (NST) and Generative Adversarial Networks (GANs). This paper examines various multi-image mixing approaches utilized in recent plant disease diagnosis research. Various problems encountered while using these approaches and their potential solutions for plant disease identification are also explored in the study.