Habiba Omran Thesis
This project presents a systematic comparative study of deep convolutional neural networks to classify cervical cell images from the public SIPaKMeD dataset into normal or abnormal classes. It evaluates custom compact CNN architectures versus state-of-the-art pretrained networks to provide insights into accuracy, efficiency, computational costs and generalizability for real-world clinical adoption of deep learning-based automated cervical cancer screening tools. Varuios pre-trained pytorch architicture models have been used here as: Mobilnet-v3 small, VGG 16, 19,and Resnet 50