Deep Learning Approach for Automated Vine Classification

Convolutional Neural Networks (CNNs) are popular with their ability to generalize, but CNNs require a relatively large amount of data to converge. This research studies the performance of different CNN models on a small and imbalanced dataset of vine leaves. The research shows that combining some methods like up-sampling, fine-tuning and voting could improve the baseline performance of different models and result in an Accuracy up to 94.6% from a single model, and 99.3% by voting