Predicting Geckler Classification from Gram Stained Smears Images for Sputum: Image Classification versus Object Detection

A Geckler classification is a criterion how the image is quality based on the number of buccal squamous epithelial (BSE) cells and leukocytes in the Gram stained smears images per $100 \times$ field for sputum. The Geckler classification then determines which of images is valuable to microscope testing for the Gram stained smears images per $1, 000 \times$ field for sputum. In this paper, we predict the Geckler classes by using two methods, image classification and object detection. For the image classification, we predict the Geckler classes from the training images for every Geckler class, without annotating BSE cells and leukocytes. On the other hand, for the object detection, we detect BSE cells and leukocytes from the training images annotated by them and determine the Geckler class by the number of them. Hence, we compare the experimental results of predicting Geckler classes obtained by the image classification with those by the object detection.