Visual Pollen Classification Using CNNs and Vision Transformers

Pollen classification is addressed by creating a new dataset of four common fruit plant pollens to fill a data gap in ecology and biotechnology. A fine-tuned YOLOv12 model is used for object detection to extract individual pollen grains from complex microscope images. This process results in two distinct datasets for training, created using different confidence thresholds to manage image quality and quantity. The work then proceeds to classify the individual pollen images using various models, including Convolutional Neural Networks and Vision Transformers, to find the most effective approach.