A multi-magnification digital colposcopy image datasets for cervical cancer screening in Nigeria.

This dataset comprises 3356 high-resolution digital colposcopy images collected from 332 women who attended cervical cancer screening at Bowen University Teaching Hospital, Ogbomoso, Nigeria, between June and July 2025. Each participant underwent comprehensive colposcopy imaging at four magnification levels (1 ×, 3 ×, 5 ×, and 7 ×) both before and after the application of 5% acetic acid solution. The acetic acid test, administered with a 45-second waiting period, enhances visualization of cervical epithelial abnormalities through characteristic acetowhite changes. All images were captured at 2048 × 1536 pixel resolution in JPEG format using medical-grade digital colposcopy equipment with integrated LED illumination. The dataset spans two diagnostic categories based on expert colposcopic interpretation: Normal (210 participants, 2134 images) and Abnormal (122 participants, 1222 images). The Abnormal category encompasses cervical lesions identified by established colposcopic features including acetowhite changes, punctation, and mosaicism. Images are organized in a patient-centric folder structure with standardized filenames encoding acquisition parameters, facilitating machine learning model development. Expert gynecologists classified and annotated the images using the Computer Vision Annotation Tool (CVAT) platform. This annotated collection enables the development and benchmarking of artificial intelligence models for automated cervical cancer detection, with particular relevance for deployment in resource-constrained healthcare environments where specialist expertise remains scarce.