Polyploidy, the variation in chromosome sets within plants, influences stomata size and density, making stomata analysis a valuable method for determining ploidy levels. Traditional microscopic analysis, nevertheless, is often labor-intensive and complex. This study explores the use of artificial intelligence for the automated classification of plant ploidy levels from stomata images, presenting a novel approach in this field. Experiments were conducted on three blackcurrant genotypes: diploid, triploid and tetraploid. Deep learning techniques were employed for stomata segmentation and classification, with performance compared to traditional machine learning algorithms, including K-Nearest Neighbors, Support Vector Machine, Random Forest and Multi-Layer Perceptron. To mitigate the impact of color variations that could lead to inflated accuracy, multiple datasets were processed to reduce the influence of color. Classification was performed not only on whole images but also on subimages containing individual stomata instances, detected using the YOLOv8 algorithm. A majority voting approach was applied to classify the entire image based on subimage classifications. ResNet152v2 achieved the highest accuracy of 0.973 on color images, although accuracy declined when the influence of color was minimized. These results underscore the significant role of color in model performance and highlight the challenges associated with achieving reliable and robust classification.