Cervical cancer is a tremendous concern worldwide, with limited or no early detection and efficient screening in low-resource settings. It plays a critical role in detecting abnormal regions such as Pap smear and colposcopy in the medical images, which is crucial for cervical cancer diagnosis. With recent advances in deep learning, we may be nearing a solution where image segmentation of cervical cancer can finally reach acceptable accuracy. This survey investigates novel deep learning algorithms used to identify cervical cancer, emphasizing segmentation methods such as CNN-based architectures, U-Net, and its derivatives. The research analyzes current datasets in-depth, evaluates essential performance indicators, and explores the issues of generalization, class imbalance, and segmentation accuracy. Furthermore, the survey identifies future research objectives to better deep learning incorporation into clinical practice, segmentation approaches, and cervical cancer therapy. It provides vital insights for academics and healthcare professionals trying to improve cervical cancer diagnosis and treatment by bridging the gap between research and real-world applications.