This paper presents an accuracy assessment of a hand-held ECG acquisition device using 2-D convolutional neural network (CNN) based multi-classification of electrocardiogram (ECG) images. A deep learning-based classification model has been proposed in this paper to anticipate the user-friendly approach to ECG assessment. Many classification models have been proposed earlier for heart disease prediction, most of which have used spectrum analysis. However, in a practical scenario, a medical expert never prefers spectrum analysis or sampled ECG signals. In some critical cases, ECG is not fully conclusive, and other cardiac parameters are also required for accurate prediction of cardiac disease. Even though ECG image data is the key parameter to predict or detect heart disease for medical experts. Therefore, an image-based ECG classification approach has been proposed in this paper to provide an ease of access methodology for patients as well as for the medical expert. The proposed CNN-based classification model is trained and tested with the Kaggle ECG image dataset on the cloud, and the validation of the proposed model is performed on a real-time patients’ dataset taken from the Sanketlife ECG acquisition device. The testing accuracy for the proposed model is only 89.6%, as the ECG data acquired from the Sanketlife device is divided into several beats manually. The validation accuracy for the proposed algorithm is 97.29%.