Brain tumours, the most common and hostile illness, have a relatively low survival rate during their most mature stage. As a result, therapeutic planning is a critical stage in raising the standard of living of sufferers. Different imaging modalities, including computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound pictures, are frequently used to assess malignancies inside the brain, lungs, liver, breast, prostate, and some other regions.These MRI images cast-off to diagnose brain tumors in particular in this work. However, the abundance of information gathered through an MRI scan creates manual categorization of tumour cells vs. non-tumor in a specified timeframe is unrealistic. However, it has numerous restrictions (e.g., consistent computable measures are only available aimed at a limited amount of photos). As an outcome, dependable and automatic classification technique is required to reduce human mortality. Due to the significant geographical and structural variety of the brain tumor's surrounding environment, automatic brain tumour categorization is indeed a challenging task. This research proposes utilizing Convolutional Neural Networks (CNN) categorization to automatically identify brain tumours. The underlying design was built using smaller kernels. The mass of a neuron is regarded as being very little. Test results reveal that CNN records get 93% with minimum complexity as contrasted to many other state-of-the-art approaches.