Enhancing the Efficiency in Water-Meter Monitoring System using Neural Networking

This paper presents a water meter monitoring system enhanced by deep learning techniques such as Artificial Neural Networks(ANN) and Residual Networks(ResNet). Accurate readings of the water meters are essential for effective billing and resource management; but the limitations of traditional analog meters and previous computational methods include precision, scalability, and adaptability. The implementation of deep learning techniques has enabled the system to improve its ability to extract features and classify them with improved accuracy, resulting in a test accuracy of 99.17%. Additionally, High dynamic range (HDR) imaging and color-invariant edge detection are briefly studied upon in order to address the challenges associated with inconsistent lighting and varying environments causing gradient errors. The proposed model utilizes transfer learning from pretrained networks and datasets, optimizing for smaller datasets, and enhancing performance.