Schizophrenia, a severe mental health disorder affecting approximately 24 million people globally (0.32% of the population), disrupts personal, social, and occupational functioning through symptoms that include hallucinations, disorganized behavior and delusions. Current diagnostic methods are time-intensive and subjective, necessitating advancements in automated detection. Electroencephalogram (EEG) data captures fluctuations in neural activity associated with human memory processes. This study introduces SCZ-MDRNet, a Multi-Scale Dilated Residual Convolution Network designed for automated diagnosis of schizophrenia from EEG signals. The approach involves segmenting EEG data by a process called windowing to expand the dataset, followed by conversion of the samples into color-mapped visualizations for input into the model. Using the IBIB-PAN dataset comprising of EEG recordings from 14 schizophrenic and 14 healthy individuals, with 19-channels, recorded at 250 Hz, SCZ-MDRNet achieves a test accuracy of 99.90%, perfect diagnostic performance with an AUC, specificity, sensitivity, and F1 score of 1. These results underscore the proposed model’s potential as a transformative tool for reliable and efficient schizophrenia diagnosis. Future work will focus on addressing limitations due to the small dataset size and incorporating more diverse datasets to improve robustness.