Abstract—This study concentrates on the development of AgroExpense-OCR, an Optical Character Recognition (OCR) system for agricultural expense receipts, employing the Vision Transformer (ViT) architecture to improve the recognition of both printed and handwritten text. Utilizing a dataset of over 15,000 systematically prepared items, experimental results demonstrate that FastViT achieves the fastest convergence and lowest Character Error Rate (CER), despite transient instability during handwritten sample training. The best character-level performance is achieved by FastViT, with a test loss of 0.03144 and Certificate of Analysis (CER) of 0.0086 using greedy decoding and 0.0091 with beam search. Evaluation results demonstrated superior performance, with printed receipts attaining an accuracy of 96.80% and handwritten receipts 92.40%, surpassing traditional CNN-RNN methodologies. The AgroExpense-OCR application was created for both mobile and web platforms, including features such as receipt scanning, automatic expense categorization, cost analysis, and budget notifications. User satisfaction assessments indicated a high level of acceptance, thereby confirming the practical applicability of the system. This research makes a significant contribution to the advancement of digital agriculture and establishes a foundation for integration with future smart farm management systems.
 
Keywords—agricultural receipts, digital agriculture, optical character recognition, smart farming, vision transformer


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).