Hybrid RGB-CNN-SVR modeling for enhancement of smartphone colorimetric sensing using a Schiff base sensor.

Although Schiff base colorimetric sensors provide simple visual detection of metal ions, reliable quantitative analysis using conventional RGB-based smartphone methods remains limited by relatively low precision. In this study, a smartphone-based hybrid deep learning framework was developed by combining convolutional neural network (CNN) features with colorimetric descriptors, such as RGB and CIE Lab* parameters, and SVR to improve the precision of Fe2+ quantification using a test-strip-based Schiff base sensor. The proposed method achieved excellent predictive performance, with a mean R2 of 0.992, while improving the limit of detection (LOD) and limit of quantification (LOQ) by factors of 1.331 and 1.466, respectively, compared with using the RGB method alone. Furthermore, this proposed approach will pave the way for portable, intelligent metal-ion analysis, enabling accurate concentration determination with only a smartphone.