In this paper (purchase required) the authors applied an open-access AI model that is routinely used in automated object recognition systems (You Only Look Once – YOLO) to honey authentication using microscopy of pollen. They created a data set comprising three well-known honey varieties (Sundarban, Litchi, and Mustard), supplemented by three sets of unidentified honey pollen images sourced from Kaggle (an open-access repository of machine learning data). They assembled a data set consisting of 3000 images representing the pollen types extracted from the known honey samples. To tackle the challenge of limited sample sizes, they employed data augmentation techniques.
They reported good statistical performance characteristics including detection accuracy, precision, recall, mAP value, and F1 score. They applied the model to Kaggle’s unknown honey pollen data sets, and reported that it correctly detected and identified these new pollens based on previous training.
Photo by David Clode on Unsplash
Comments