PERBANDINGAN CNN, RESNET50, DAN VISION TRANSFORMER UNTUK KLASIFIKASI KANKER PAYUDARA BERBASIS WEB

Stella Juventia Grace (Universitas Muhammadiyah Surakarta)
Dedi Gunawan (Universitas Muhammadiyah Surakarta)

RABIT: Jurnal Teknologi dan Sistem Informasi Univrab
Vol 10 No 2 (2025): Juli

This research aims to compare three deep learning algorithm-based image processing models, namely CNN, ResNet50, and Vision Transformer (ViT), in classifying breast cancer based on mammography images. The CBIS-DDSM dataset from Kaggle was used and processed through pre-processing steps such as data cleaning, image resizing, normalization, augmentation, and data splitting into training and testing sets. The models were evaluated using a 5-Fold Cross Validation scheme to ensure performance stability. The results show that ResNet50 achieved the highest accuracy of 97%, followed by CNN at 92%, and Vision Transformer at 71%. All three models were implemented into a web application using Flask to support the automatic diagnosis process. These findings are expected to help develop a faster and more accurate breast cancer detection system for medical professionals.