Classifying motor imagery electroencephalograph (MI-EEG) signals remains one of the central challenges in brain-computer interface (BCI) systems. Despite numerous advancements involving signal processing techniques and artificial intelligence, enhancing the classification accuracy of MI-EEG signals remains constrained. We have developed a CNN-based model inspired by EEGNet, designed with lower computational complexity specifically for classifying the motor imagery task. We used the raw data from the BCI competition IV-2a dataset without any preprocessing to train the model. Simulation results show that our model effectively classifies the MI-EEG signal. Finally, we conducted a comparative analysis between the proposed model and the EEGNet model. Our proposed model achieved a classification accuracy of 79.99%, surpassing the EEGNet model by approximately 4%. The proposed model is simple, more robust and accurate for real-world applications.