Performance Evaluation of Deep Learning Models for Pap-Smear Image Classification

Cervical cancer is the fourth most common cancer in women worldwide. In this study, we used the open dataset SIPAKMED dataset. This dataset contains 4049 pap-smear images, and the images are categorized into five categories: Superficial-intermediate, parabasal, coilocytotic, dyskeratotic and metaplastic. In the classification phase, 60% of the images in this dataset were used as training dataset, 20% as test dataset and the remaining 20% as validation dataset. The number of epochs for the models was chosen as 30. The success rates obtained with the VGG19, InceptionV3 and SSCNN models are 93%, 91% and 89% respectively.