A brain tumor is a condition where growth of cells within or around the brain becomes abnormal. Tumors can be benign or malignant. They are categorized as primary which originate in the brain or secondary which spreads from other body parts to the brain. Common tumor types include gliomas, meningiomas, pituitary tumors, schwannomas, and medulloblastomas. Symptoms vary but may include headaches, seizures, or cognitive changes. Treatment depends on the tumor's type, size, and location. The dataset used in this paper only consists of gliomas, meningiomas, pituitary tumors and no-tumor class used as the baseline category. It does not specify if they are benign or malignant, primary or secondary. The dataset is gathered from Kaggle, consisting of 5712 training and 1311 test data of brain MRI images distributed among 4 classes. Deep learning models (DenseNet121, MobileNetv2, ResNet50, DenseNet169, DenseNet201) were used with model freezing and parameter tuning to achieve better results, where DenseNet169 gave the highest accuracy of 96.19%. This system can be further developed by carefully tuning the parameters and we intend to expand and merge more datasets together in the future.