Precisely identifying and maintaining electricity meters is essential to equitable invoicing, effective energy allocation, and efficient resource management in utility systems. Accurate meter authentication is crucial during replacement initiatives, and electricity theft is becoming a more serious threat to the efficient supply of electricity. A refined Convolutional Neural Network classifier was designed to distinguish between the Electric Meter and Non-Electric Meter classes. Raw data was preprocessed by applying image enhancement, skew correction, and dimensionality reduction techniques to a dataset comprising 5121 images across two classes. Initially, few established models: MobileNetV2, EfficientNetB1, InceptionV3 and DenseNet121 were applied, yielding accuracies of 97.70%, 99.60%, 90.90% and 97.40% respectively whereas a refined VGG16 achieves the highest accuracy at 99.70%. Performance metrics and an ablation study compare the models, showcasing the superior classification and detection capabilities of the suggested model for Electric Meters. Notably, this model was integrated into a mobile application for practical use. This study lays the groundwork for future developments in data-driven decision-making and smart grid technologies in addition to addressing current utility management issues.