MV-AINet: Multiview-Attention Integration Network for Age Recognition of the Pericarpium Citri Reticulatae

Pericarpium citri reticulatae (PCR) has varying medicinal values from different ages. In the application of existing deep learning methods to PCR recognition, they only focused on the features of a single view. Both the surface and internal capsule features of PCR have important age characteristics. This article investigates the automatic age recognition of the PCR based on multiview learning. In this article, a multiview-attention integration network for identifying the age of PCR is proposed. In the feature extraction process, dual attention convolutional (DAC) module is built to extract features of the two views of PCR. In the view integrate process, view attention integration module is constructed, which can adaptively learn positive relations between views. Experimental results showed that, compared to existing baseline methods, MV-AINet achieved improvements of 7.08%, 6.15%, 7.83%, and 6.99% in accuracy, precision, recall, and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${F}1$ </tex-math></inline-formula> score, respectively, ultimately achieving an accuracy of 98.23% on the PCR dataset. Although the algorithm performs excellently in standard experiments, its deployment on resource-constrained embedded devices may require further optimization. This study provides an efficient and intuitive solution for the intelligent development of the PCR industry and offers a referenceable approach to addressing multiview learning problems.