Cervical cancer is a major global health concern and the fourth leading cause of cancer in women. It is one of the most preventable cancers if detected at an early stage. However, manual screening of Pap tests is time-consuming, inefficient, and prone to errors. This study aims to automate cervical cell classification using DenseNet201 for feature extraction, with a spatial attention module (SAM) added to focus on important features. The proposed model improves the representation of spatial characteristics, leading to a more accurate classification. To evaluate the proposed model, we use the Mendeley LBC and the Pomeranian dataset. The model achieves an accuracy of 98.95% for multiclass classification and 100% for binary classification under 5-fold cross-validation on the Mendeley dataset. These results demonstrate the potential of the proposed model to assist cytopathologists in improving screening and reducing human error in cervical cancer diagnosis.