Deep Learning-Integrated Agarose-Coated Micro-Loop Fiber Resonator for Relative Humidity Measurement

This work presents a fiber-optic sensor for relative humidity measurement, integrated with a deep learning model to address performance limitations of conventional fiber sensors. The proposed micro-loop resonator was coated with 0.5% agarose, exhibiting a sensitivity of 0.2037 dB/%RH within the 35%RH–85%RH range. The sensor’s time stability and temperature dependence were also examined, showing a maximum power fluctuation of 0.0993 dB. Spectral responses from the coated micro-loop were recorded and converted into Parula heatmaps, then gray-scaled, resized, and compressed to generate a dataset of approximately 5000 samples for training and testing a convolutional neural network (CNN). The integrated CNN model achieved an accuracy of 98%, demonstrating the feasibility of combining optical sensing and deep learning for enhanced humidity measurement.