Smart Phone-Based Intelligent Invoice Classification Method Using Deep Learning

Conventional invoice reimbursement methods often require excessive human resources. Intelligent invoice reimbursement (IIR) has been received strongly attentions since it can improve process efficiency and also save manpower cost. Among these procedures, invoice classification is considered one of most important steps. This paper aims to improve IIR performance capabilities by proposing a deep learning based method to intelligently classify invoices photographed by smart phones. The end-to-end method can directly identify the category of input various invoice images in a short time. Experimental results indicate that the proposed method can achieve both high classification accuracy 99.05 % and fast running speed 18.16 seconds for 105 invoice images.