<![CDATA[This Advancements in computational technology have driven the development of Deep Learning, particularly Convolutional Neural Networks (CNN), in the classification and recognition of digital images. This research focuses on the classification of MRI brain tumor images using the VGG-16 architecture. The primary challenges include gradient vanishing and overfitting due to a small dataset. The objective of the study is to evaluate the performance of the model with various data augmentation techniques and to assess the impact of different dataset compositions (90:10 and 70:30) for training and testing. Two model configurations are used: Model A with 4096 neurons and Model B with 128 and 64 neurons in the first two Dense layers, respectively. The tested augmentation techniques include rotation, flip, Zoom , and their combinations. The results indicate that rotation and Zoom augmentations provide the best performance for both models and dataset compositions. Model A (90:10) achieved an accuracy of 96% with rotation and 92% with Zoom, while Model B (90:10) achieved 94% with rotation and 98% with Zoom. For the 70:30 composition, Model A achieved 94% (rotation) and 90% (Zoom ), while Model B achieved 95% (rotation) and 96% (Zoom ). This research provides valuable insights into optimizing VGG-16 architecture for brain tumor classification using limited datasets.]]>
Copyrights © 2026
