A key medical diagnostic test is the detection of brain tumors, and prompt and accurate results are essential for efficient treatment planning. In order to develop a reliable convolution neural network (CNN) model for automating the diagnosis of brain tumors using magnetic resonance imaging (MRI) data, this research focuses on leveraging deep learning capabilities. The main objective is to classify brain tumors with high accuracy and efficiency using a transfer learning system based on well-known CNN architectures as VGG 16, INCEPT v3, and RESNET 50. The goal of this project is to create a successful constitutional network-based floral categorization method using transfer learning. This paper compares network initialization models to transfer learning models based on VGG 19 v3, inception 16v3, and ResNet50. The findings demonstrate that over-fitting and local optimal issues in deep convolution networks can be successfully avoided by transfer learning. The precision with which tumors are detected has obviously increased, and their robustness and generalization are far superior to that achieved by traditional methods.