Brain tumors are currently the biggest cause of mortality throughout the globe. This pattern has been seen in recent years. A brain tumor is characterized by abnormal cell proliferation in the brain. The purpose of this research is to develop a dependable brain tumor segmentation model based on the GoogLeNet architecture and compare its performance to AlexNet and VGGNet. Data pre-processing, Google model training, and testing on a preset dataset are all essential components of the approach. A comparison with AlexNet and VGGNet illustrates our proposed model's utility and specificity in brain tumor segmentation. The results emphasize GoogLeNet's utility by indicating potential breakthroughs in early tumor identification and classification. The study focuses on the utilization of cutting-edge medical imaging technologies, such as GoogleNet and the Walrus Optimization Algorithm, to increase diagnosis accuracy and treatment choices for brain tumors.