Comparison of OCR APIs for Digit Recognition in Seven Segment Display Images

• Jonathan R. da Silva IFES
• Leandro C. Resendo IFES
• Jefferson O. Andrade IFES
• Karin S. Komati IFES

DOI: https://doi.org/10.5753/encompif.2021.15948

Abstract

Cloud computing platforms make text recognition technology accessible. However, the most suitable solution for a given application is not always evident. This paper evaluated five different text recognition solutions: AWS Rekognition, Microsoft Azure, Cloudmersive, Google OCR, and OCRSpace. A database of images of seven-segment displays in electricity meters, the “YUVA EB Dataset”, was used. There was no pre-processing to improve image quality, to improve lighting or to eliminate noise. Google Cloud showed better results, hit-ting 100 results of the 169 input images, with an accuracy of 86.5 % considering the 965 digits. The results obtained suggest that the use of the solutions offered commercially are not suitable for use in production without a previous stage of pre-processing of the images.

Keywords: Digit recognition, Cloud computing, Seven segment display

References

Anda, F., Lillis, D., Le-Khac, N., and Scanlon, M. (2018). Evaluating automated facial age estimation techniques for digital forensics. In Proceedings of 2018 IEEE Securityand Privacy Workshops (SPW), pages 129–139. IEEE.

Bonačić, I., Herman, T., Krznar, T., Mangić, E., Molnar, G., & Čupić, (2015). Optical character recognition of seven-segment display digits using neural networks. In 32st International Convention on Information and Communication Technology, Electronics and Microelectronics, volume 3.

Finnegan, E., Villarroel, M., Velardo, C., and Tarassenko, L. (2019). Automated method for detecting and reading seven-segment digits from images of blood glucose metres and blood pressure monitors. Journal of Medical Engineering & Technology, 43(6):341–355.

Kanagarathinam, K. and Sekar, K. (2019). Text detection and recognition in raw image dataset of seven segment digital energy meter display. Energy Reports, 5:842–852.

Torres, W., van den Brand, M. G., and Serebrenik, A. (2020). Suitability of optical character recognition (ocr) for multi-domain model management. In International Conference on Systems Modelling and Management, pages 149–162. Springer.