This paper presents a faster region-based convolutional neural network (FRCNN) architecture based on regions of interest (ROI) for annotated image classification in order to detect ovarian cancer. Three categories epithelial, germ cell, and stroma cell are used in the process to classify input photos. The FRCNN framework is used for annotation after ROI is used for preprocessing and segmentation. By contrasting hand annotation features with FRCNN-trained features for region-based classification, this method mathematically shows that machine learning-based classification outperforms human techniques in terms of accuracy. To improve classification performance, particularly in datasets with higher indexing complexity, the FRCNN output is additionally subjected to region-based training using an ensemble classifier that combines Gaussian Naïve Bayes (NB), support vector regression (SVR), and support vector classifier (SVC). In order to effectively detect ovarian cancer, simulations verify the suggested approach by generating extremely accurate input image segmentation and classification. The possibilities of FRCNN and ensemble learning for reliable and precise illness diagnosis are highlighted by this architecture.