Publications

Abstract : Cervical intraepithelial neoplasia (CIN) is a major problem women face worldwide. The classic Pap smear analysis (Papanicolaou) is a suitable method for assessing cell images to diagnose cervical disorders. Many computer vision algorithms may be utilized to identify the cancerous and non-cancerous pap smear cell images. The majority of existing research focuses on binary classification techniques that use different methods. However, they have intrinsic difficulties with the excision of minor features and exact categorization. We propose a novel approach for performing multiclass classification of cervical cells with optimal feature extraction, minimal parameters, and less computing power than competing models. The implementation of ConvNet with the Transfer Learning approach validates significant cancer cell diagnosis. The suggested binary and multiclass classification techniques obtained 99.3% and 97.3% accuracy results, respectively, on the dataset.