In　ovarian　cancer　image　classification,　accurate　advance　prediction　is　essential　for　diagnosis　and　treatment.　A　model　combining　EfficientNet-B0　and　Generalized　Mean　(GeM)　Pooling　is　proposed　in　this　paper.　The　main　goal　of　the　model　is　to　capture　tiny　features　and　changes　in　ovarian　cancer　images　and　use　these　features　to　divide　the　images　into　five　subtypes,　thus　enabling　advanced　prediction　and　accurate　diagnosis　of　ovarian　cancer.　The　experimental　results　highlight　the　variation　in　the　model＇s　performance　at　different　epochs,　clearly　showing　when　the　model　is　at　its　best.　This　comprehensive　approach　provides　a　comprehensive　solution　for　the　classification　of　ovarian　cancer　subtypes　and　demonstrates　the　potential　to　integrate　different　neural　network　architectures,　further　deepening　our　understanding　of　ovarian　cancer　images.　Thus,　the　study　makes　an　important　contribution　to　the　field　of　ovarian　cancer　subtype　classification,　validating　the　effectiveness　of　integrating　GeM　to　improve　model　performance.　©2024　IEEE.