Ovarian cancer is a proliferation of ovarian cells. The cells multiply rapidly, allowing them to penetrate and destroy healthy human tissue. By using microscopic examination, pathologists can diagnose five common histotypes - Endometrioid Carcinoma (EC), Clear-cell Carcinoma (CC), Low-Grade Serous Carcinoma (LGSC), High-Grade Serous Carcinoma (HGSC), and Mucinous Carcinoma (MC). In this work, ovarian carcinoma and its subtypes are detected using MobileNet V2 and ResNet 50. ResNet-50 is chosen for tasks requiring deeper architectures and higher accuracy, while MobileNetV2 is preferred for resource-constrained environments due to its efficiency and speed. This is evaluated on the publicly available histopathological dataset. Generalized Mean (GeM) pooling is used for splitting the Whole Slide Image (WSI) images into respective tiles. These tiles are given as an input for further processing. Feature extraction methods like grayscale features, color histograms, and contouring are performed on these tiles. Each tile is analyzed. These tiles are given as input for training, testing, and validation using Residual Network 50 (ResNet 50) and MobileNet V2 architectures. The output obtained will be the image ID, tile number, and the histotype detected. As a result, this model obtained a training, testing and validation accuracy of 93.85%, 96.01%, and 97.11% respectively using ResNet 50 and training, testing, validation accuracy of 83.85%, 87.69%, 86.75% respectively using MobileNet V2. Therefore, it can be concluded that ResNet 50 gives better performance than MobileNet V2. Identifying ovarian cancer is critically important due to the high mortality rate associated with late diagnosis, the absence of specific signs as well as the absence of efficient screening methods. Earlier detection significantly improves survival rates, and ongoing investigation into ovarian cancer enhances our understanding of the disease, leading to better treatment, prevention, and healthcare outcomes.