Analysis of Various CNN Models for Locating Keratin Pearls in Photomicrographs

Abstract

Worldwide, cancer occupies the second position in the list of diseases associated with the high mortality rate. Cancer is the condition, where in the body immune system will not be having any control over the population of the cells and their division process. Oral cancer is the cancer associated with the oral cavity or the mouth area. The detection of cancer from the photomicrographs is done manually by the pathologists and oncologists. This paper tries to automate the detection process by locating the keratin pearls which are rose like patterns found in the photomicrographs of malignant tissues. This is achieved through the Tensorflow object detection API freely available through the github repository. We have used four different models used for the coco data set. The analysis of the same is presented here.