Heart disease is now the leading cause of mortality globally, in both developed and developing countries. Lower death rates can be achieved with early diagnosis of heart problems and ongoing clinical supervision by experts. A disproportionately large number of deaths worldwide are caused by cardiovascular conditions, which have a high risk of natural death. An effective early detection technique is necessary to reduce the death rate from cardiovascular diseases (CVDs). Electrocardiograms (ECGs) are used to understand a range of heart conditions in patients. This research directly impacts improving healthcare at a cheaper cost and saving lives as healthcare and health insurance costs increase globally. Deep learning methods were employed in this study to predict the four primary cardiac abnormalities—aberrant heartbeat, myocardial infarction (MI), history of myocardial infarction, and normal person classes—from the publicly available ECG picture dataset of cardiac patients. We built the system for this project using the MobileNet architecture and succeeded in obtaining a validation accuracy of 91.00% and a training accuracy of 97.34%. Consequently, the suggested MobileNet architecture model can accurately diagnose cardiovascular diseases and may also be used to generate features for traditional machine learning classifiers. Medical experts can utilize the proposed MobileNet architecture model to identify cardiac diseases using ECG images, replacing the manual method that yields inaccurate and time-consuming findings.