Ensemble Transfer Learning for Automated Gauge Reading Detection and Prediction

Abstract

Pressure gauges and automatic reading methods for pointer gauges were the root causes of the problems that have motivated the development of an ensemble transfer learning strategy. The study suggests that to effectively generate and predict the present measurements of the gauge, it is necessary to use an ensemble learning method that incorporates transfer learning framework designs such as InceptionResnetV2 or DenseNet 201. The suggested methodology involves integrating the given data with ensemble model architectures and qualifying InceptionResnetV2 and Dense Net 201 models to forecast the present gauge value. The main focus of the proposed method is to develop a final angle with a clear reading, as well as improve the picture through techniques like enhancing its shape, size, and resolution. Image processing approaches help to achieve these objectives. The ensemble model achieved an accuracy of 98.34%, whereas the InceptionResNetV2 model experienced a loss of 8.34% in contrast stretching.