Brain tumor is one of the most hazardous disease that leads a man to gradual death. To ensure proper and effective treatment, this is very important to detect the brain tumor and predict this as cancerous or non-cancerous. Radiologists have shown interest to detect brain tumor and its category analyzing the MRI (Magnetic Resonance Image) of brain. This detection and classification task seems to be challenging because of different size, location and behavior of brain tumors. Deep learning based classifiers extract features from MRI and helps to diagnose brain tumor with the help of computer aided diagnosis system. In this paper, we have experimented this classification task on a publicly available dataset using transfer learning approach in InceptionV3 and DenseNet201 model. Data augmentation technique is performed to enrich the dataset for achieving a good classification result an to avoid over fitting.“Brain-DeepNet” a deep convolutional neural network has been proposed where six convolution layers are densely connected and extract features from dense layers. These dense layers extract features and all features are passed to a fully connected layer. Dense network extract features more efficiently from brain MRI. This work is experimented on MRI as MRI provides more details of cell structure and functions. Our proposed model has shown approximately 96.3% classification accuracy to differentiate among the three types of brain tumors most commonly encountered Glioma, meningioma, and pituitary. This model outperforms the classification performance in comparison with the pretrained models.