Investigating the Impact of Input Resolution on Object Detection Performance Using EfficientDet for Signature Detection in Document Images

Authors

  • Ensophea Toch School of Graduate Studies, Paragon International University, Phnom Penh, Cambodia Author
  • Pakrigna Long Department of Applied Mathematics and Statistics, Institute of Technology of Cambodia, Phnom Penh, Cambodia Author

Keywords:

signature detection, object detection, EfficientDet, BiFPN, input resolution, document image analysis

Abstract

Automated signature detection is a critical component of intelligent document processing systems, supporting document authentication and fraud detection. This paper presents a controlled empirical investigation of the effect of input resolution on handwritten signature detection using EfficientDet-D0. Three resolutions—512×512, 640×640, and 768×768—were evaluated under identical training conditions on a self-collected corpus of 1,000 annotated document images. Detection performance, measured by mAP@0.5, follows a non-monotonic trend: it improves from 0.9000 at 512×512 to a peak of 0.9261 at 640×640, then declines slightly to 0.9233 at 768×768, even though the 768×768 configuration achieves the lowest validation loss. We attribute this pattern to interacting effects of BiFPN feature granularity, anchor-scale alignment, and optimization complexity, and we discuss the scope and limitations of a single-architecture, single-run study. The results indicate that input resolution is a domain-specific hyperparameter that should be empirically tuned rather than inherited from benchmark conventions.

References

[1] M. Blumenstein, X. Y. Liu, and B. Verma, “A modified direction feature for cursive character recognition,” Pattern Recognition, vol. 40, no. 2, pp. 376–388, 2007. [Online]. Available: https://doi.org/10.1109/IJCNN.2004.1381140

[2] T.-Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature pyramid networks for object detection,” CVPR, 2017. [Online]. Available: https://arxiv.org/abs/1612.03144

[3] M. Tan, R. Pang, and Q. V. Le, “EfficientDet: Scalable and efficient object detection,” CVPR, 2020. [Online]. Available: https://arxiv.org/abs/1911.09070

[4] T.-Y. Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Dollár, and C. L. Zitnick, “Microsoft COCO: Common objects in context,” ECCV, 2014. [Online]. Available: https://doi.org/10.1007/978-3-319-10602-1_48

[5] H. Yuan et al., “Online handwritten signature verification based on temporal-spatial graph attention transformer,” arXiv preprint, 2025. [Online]. Available: https://arxiv.org/abs/2510.19321

[6] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards real-time object detection with region proposal networks,” IEEE TPAMI, vol. 39, no. 6, pp. 1137–1149, 2017. [Online]. Available: https://arxiv.org/abs/1506.01497

[7] N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko, “End-to-end object detection with transformers,” ECCV, 2020. [Online]. Available: https://doi.org/10.1007/978-3-030-58452-8_13

[8] W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.-Y. Fu, and A. C. Berg, “SSD: Single shot multibox detector,” ECCV, 2016. [Online]. Available: https://doi.org/10.1007/978-3-319-46448-0_2

[9] X. Zhu, W. Su, L. Lu, B. Li, X. Wang, and J. Dai, “Deformable DETR: Deformable transformers for end-to-end object detection,” ICLR, 2021. [Online]. Available: https://openreview.net/forum?id=gZ9hCDWe6ke

[10] K. Yan, Y. Zhang, H. Tang, C. Ren, J. Zhang, G. Wang, and H. Wang, “Signature detection, restoration, and verification: A novel Chinese document signature forgery detection benchmark,” CVPRW, 2022, pp. 5163–5172. [Online]. Available: https://doi.org/10.1109/CVPRW56347.2022.00564

[11] Y. Zhang, J. Zhang, K. Yan, et al., “Handwritten Chinese signature detection with simple Copy–Paste augmentation on power plants technical documents,” Service Oriented Computing and Applications, vol. 17, pp. 293–302, 2023. [Online]. Available: https://doi.org/10.1007/s11761-023-00374-8

[12] G.-S. Xia, X. Bai, J. Ding, Z. Zhu, S. Belongie, J. Luo, M. Datcu, M. Pelillo, and L. Zhang, “DOTA: A large-scale dataset for object detection in aerial images,” CVPR, 2018, pp. 3974–3983. [Online]. Available: https://doi.org/10.1109/CVPR.2018.00418

[13] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” CVPR, 2016, pp. 770–778. [Online]. Available: https://doi.org/10.1109/CVPR.2016.90

[14] P. Zhu, L. Wen, D. Du, X. Bian, H. Ling, Q. Hu, and Q. Nie, “Detection and tracking meet drones challenge,” IEEE TPAMI, vol. 44, no. 11, pp. 7380–7399, 2021. [Online]. Available: https://doi.org/10.1109/TPAMI.2021.3119563

[15] I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” ICLR, 2019. [Online]. Available: https://openreview.net/forum?id=Bkg6RiCqY7

[16] Z. Wang et al., “Advances in offline handwritten signature recognition research: A review,” IEEE Access, 2023. [Online]. Available: https://doi.org/10.1109/ACCESS.2023.3326471

[17] G. K. Pandey et al., “Offline signature verification: An extensive survey,” ICSADL, 2025. [Online]. Available: https://ieeexplore.ieee.org/document/10933339

Downloads

Published

2026-11-02

Issue

Vol. 1 No. 1 (2026)

Section

Articles

License

Copyright (c) 2026 Journal of Applied Mathematics, Informatics and Digital Arts

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

All articles published in this journal are distributed under the terms of the Creative Commons Attribution 4.0 International License (CC BY 4.0).

Under this license, users are free to:

  • Share — copy and redistribute the material in any medium or format
  • Adapt — remix, transform, and build upon the material for any purpose, including commercial use

Under the following conditions:

  • Attribution — Appropriate credit must be given to the original authors and source, a link to the license must be provided, and any changes made must be indicated. Attribution must not suggest endorsement by the authors or the journal.

Author Rights

Authors retain full copyright of their work and grant the journal the right of first publication. After publication, authors may:

  • Archive the published version in institutional repositories
  • Share and reuse their work with proper attribution

Open Access Statement

This journal follows a Diamond Open Access model, meaning:

  • No publication fees (APCs) are charged to authors
  • No subscription fees are charged to readers
  • All content is freely accessible immediately upon publication

How to Cite

Investigating the Impact of Input Resolution on Object Detection Performance Using EfficientDet for Signature Detection in Document Images. (2026). Journal of Applied Mathematics, Informatics and Digital Arts, 1(1), 8-12. https://ojs.paragoniu.edu.kh/index.php/pjournal/article/view/22