Depression is recognized as a serious and widespread illness today, yet its distinct symptoms remain obscured. Given the extensive time required for the manual monitoring of long-term electroencephalogram (EEG) recordings for depression, the exploration of automated depression identification is essential. This research presents a method for the automatic detection of depression using EEG. The EEG signal’s noise was initially attenuated using a bandpass filter ranging from 0.1 Hz 70 Hz, and power line artifacts were eradicated with 50 Hz notch filter. The EEG recordings were converted into 4-s epochs. In the initial phase of our two processes, the EEG epochs were restructured for the input of the convolutional neural network (CNN). In the end procedure, the epochs were converted into heatmap images to facilitate the automatic extraction of features from the inputs and classify them as “Major Depressive Disorder” or “Healthy” patients. The proposed techniques were evaluated utilizing a publically accessible dataset, resulting in an accuracy of 92.54% for our initial method and 98.50% for our final method when heatmap images were generated.