Cervical cancer is one of the most frequent diseases in women, and it is a leading cause of death in many underdeveloped nations. Cervical lesions are diagnosed with a pap smear test or acetic acid visual examination (staining). Digital colposcopy, a low-cost screening procedure, produces painless and quick results. As a result, automating cervical cancer screening using colposcopy pictures will be very beneficial in saving many lives. Many automated approaches using computer vision and machine learning in cervical screening have garnered prominence recently, opening the door for cervical cancer diagnosis. However, the majority of the approaches depend exclusively on cervical detection and segmentation annotation. The purpose of this research is to present the Faster Small-Object Detection Neural Networks (FSOD-GAN) to address cervical screening and cancer detection utilising digital colposcopy pictures. The suggested method uses a Faster Region-Based Convolutional Neural Network (FR-CNN) to identify cervical spots and conducts hierarchical multiclass classification of three kinds of cervical cancer lesions. Experimentation was carried out using colposcopy data from publicly accessible sources on 1,993 patients with three cervical classifications, and the suggested technique had 99% accuracy in identifying cervical cancer stages.