Applications of Two Separate Methods to Deal with a Small Dataset and a High Risk of Generalization

Dimensionality reduction means obtaining additional training data by supplementing this study with supervised learning to match input variables. A training dataset is used due to a lack of training examples in some domains and avoiding the fitting problem. A Convolution Neural Network (GAN) to augment a classification model for computer vision has shown to improve prediction performance. Using the augmented neural network, this study analyses whether transfer processing within the same GAN will improve prediction performance even further. This proposed a method for assessing the effect of data augmentation in the processing of microscopic charcoal images. Morphological transformations and sub-images were various steps to data augmentation that have been investigated. A post-Inception-v3 framework is also applied to educate a barbecue species classifier using the data available. The best value was observed using the micro methodology, which had an overall precision of 99.36%