Brain tumors are a serious medical disorder that are characterized by the unchecked growth of abnormal cells inside the brain. Image processing has advanced significantly in the context of biomedical applications, especially since deep learning techniques became available. Brain tumor detection and classification constitute vital areas of research, often leveraging Convolutional Neural Networks (CNNs). But current approaches are too complex, with many parameters contributing to their complexity. This results in slow execution times and high system requirements when implementing them. We introduce a revolutionary architecture in this work that is based on the Region-based Convolutional Neural Network (RCNN) technique for both tumor type detection and brain tumor classification. We evaluate our methods using publicly available datasets from Fig share (2017) and Kaggle (2020). Our main goal is to use a low-complexity framework to expedite the execution time of traditional RCNN architectures, offering a practical remedy for the investigation of brain tumors. First, with an amazing accuracy of 98.212%, we employ a Two Channel CNN, which is renowned for its simplicity, to discriminate between MRI samples of a healthy tumor and glioma. Then, an RCNN framework is employed as the feature extractor to find tumor locations in Glioma MRI samples that were previously classified in the first stage. Bounding boxes are used to define tumor regions that have been detected. Moreover, our methodology is expanded to include two new forms of tumors: pituitary tumors and meningioma. The proposed approach demonstrates substantially reduced execution times compared to current designs, resulting in an average degree of confidence of 98.829%.