Comparative Analysis of Text Detection Accuracy on Financial Documents Using CTPN and EAST
DOI:
https://doi.org/10.22303/csrid-.18.2.2026.274-286Keywords:
CTPN, EAST, Text Detection, Financial OCR, Deep LearningAbstract
Automation of data extraction in financial documents is an urgent necessity for MYO Adventures to mitigate the risks of manual entry errors and operational inefficiencies. This study aims to evaluate the performance of Deep Computer Vision modules within an automated validation system through a comparative analysis between Connectionist Text Proposal Network (CTPN) and Efficient and Accurate Scene Text Detector (EAST) algorithms. The research methodology employs a quantitative experimental approach, testing both models against invoice and receipt datasets characterized by physical distortions and dense tabular layouts. Evaluation focuses on inference time metrics and text localization effectiveness prior to the Optical Character Recognition (OCR) stage. The results reveal a significant performance disparity, where the EAST algorithm recorded an average detection time of 565 ms, substantially more efficient than CTPN which required an extreme 12,046 ms. In terms of coverage accuracy, EAST demonstrated high robustness by successfully recognizing 74 valid text items, in contrast to CTPN which yielded only 14 readable items. The poor performance of CTPN is analyzed as a consequence of the limitations of its vertical anchor mechanism and Recurrent Neural Network (RNN) in handling the dense spacing variations of invoice tables, whereas the Fully Convolutional Network (FCN) architecture of EAST proved to be more adaptive. In conclusion, anchor-free methods like EAST possess superior reliability and computational efficiency, making them the most viable solution for implementing real-time financial validation systems on MYO Adventures' web-based platform.
References
Long, S., He, X., & Yao, C. (2020). Scene text detection and recognition: The deep learning era. International Journal of Computer Vision, 129(1), 161-184.
Reed, M., Cook, S., & Morgan, K. (2023). Hybrid CNN LSTM framework for OCR-based text extraction from receipts. Research Paper.
Soni, V. K., Shukla, V., Tandan, S. R., Pimpalkar, A., Nema, N. K., & Naik, M. (2025). Performance evaluation of efficient and accurate text detection and recognition in natural scenes images using EAST and OCR fusion. International Journal of Advanced Computer Science and Applications, 16(1).
Liu, R., Shi, Y., Tang, X., & Liu, X. (2025). Advanced dense text detection in graded examinations leveraging Chinese character components. Applied Sciences, 15(4), 1818. https://doi.org/10.3390/app15041818
Akdoğan, A., & Kurt, M. (2024). ExTTNet: A deep learning algorithm for extracting table texts from invoice images. arXiv preprint arXiv:2402.02246.
Liao, M., Wan, Z., Yao, C., Chen, K., & Bai, X. (2020). Real-time scene text detection with differentiable binarization. Proceedings of the AAAI Conference on Artificial Intelligence, 34(07), 11474-11481.
Sayallar, Ç., Sayar, A., & Babalık, N. (2023). An OCR engine for printed receipt images using deep learning techniques. International Journal of Advanced Computer Science and Applications, 14(2).
Liu, M., Li, B., & Zhang, W. (2022). Research on small acceptance domain text detection algorithm based on attention mechanism and hybrid feature pyramid. Electronics, 11(21), 3559. https://doi.org/10.3390/electronics11213559
Amari, A., Lahmar, A., Zormati, M. A., Makni, M., & Koubaa, F. (2025). An efficient deep learning-based approach to automating invoice document validation. arXiv preprint arXiv:2503.12267.
Kumar, Y. J. P., Kumari, C. S., Lakshmi, P. B., Balakrishna, G., & Rao, P. P. C. (2024). Text extraction and detection from images using CRNN and security algorithm. International Journal on Science and Technology, 29(1).
Gunaydin, E., Gencturk, B., Ergen, C., & Koklu, M. (2023). Digitization and archiving of company invoices using deep learning and text recognition-processing techniques. Intelligent Methods in Engineering Sciences, 2(4), 90-101.
Kathiravan, M., Mohan, A., Vijayakumar, M., Manikandan, M., Fernandez, T. F., & Arumugam, S. S. (2025). Enhanced textual data reconstruction from scanned receipts using normalized cross-correlation and deep learning-based recognition with superior analytical robustness and computational efficacy. International Journal of Computational and Experimental Science and Engineering, 11(3), 4713-4721.
Abdallah, A., Kasem, M. S., Elkasaby, M., Mounis, M., Mahmoud, M., & Elbendary, Y. (2025). ReceiptSense: Beyond traditional OCR - A dataset for receipt understanding. arXiv preprint arXiv:2406.04493.
Jin, Y., Jin-Chun, K. C., Wang, Y., & Ke, K. Z. (2025). Multi-stage field extraction of financial documents with OCR and compact vision-language models. arXiv preprint arXiv:2510.23066.
Huang, Y., Lv, T., Cui, L., Lu, Y., & Wei, F. (2022). LayoutLMv3: Pre-training for document AI with unified text and image masking. Proceedings of the 30th ACM International Conference on Multimedia, 4549-4559.
Sutriawan, Kurniawan, A., & Rosyid, R. (2024). Improving the quality of optical character recognition (OCR) based on neural network with the image enhancement process. Scientific Journal of Computers Sciences and Informatics, 1(1).
Liu, L. (2024). Construction of financial bill recognition model based on deep learning. Journal of Combinatorial Mathematics and Combinatorial Computing, 120, 253-264.
Long, S., Ruan, J., Zhang, W., He, X., Wu, W., & Yao, C. (2018). TextSnake: A flexible representation for detecting text of arbitrary shapes. Proceedings of the European Conference on Computer Vision (ECCV). arXiv preprint arXiv:1807.01544.
Anvari, Z., & Athitsos, V. (2022). A survey on deep learning based document image enhancement. arXiv preprint arXiv:2112.02719.
