Article Detail

Ahmet İLİSULU*, Yakup KUTLU

Keywords

Deep Learning YOLO Sign Language Classification

Doi : 10.5281/zenodo.18664511

Abstract

We actually undertook this project based on a very simple need: to alleviate some of the difficulties hearing-impaired individuals experience when expressing themselves. While thinking about how to make TID more understandable, it made sense to incorporate artificial intelligence. YOLO models are already well known for their speed and practicality, so we went straight for them. We prepared a total dataset of 9,534 images. We added not only the basic signs we knew but also signs like “Hurry up” and “Help me,” which would be useful in emergencies in daily life. Because these are the things that are actually needed in real life. YOLOv11s was very fast but a bit weak in terms of accuracy. So it works, but not quite at the level we wanted. YOLOv8m was great in terms of accuracy, but it really took its toll on the system; it was slow and training took a long time. YOLOv11m, on the other hand, struck the perfect balance. It wasn't too slow or too cumbersome. It delivered very clean results with 0.993 mAP50 and 0.991 F1-Score. In short, we decided that YOLOv11m is the most logical choice for a real-time application. It's both fast and sufficiently accurate.

References

1. Li, R., & Yang, J. (2018). Improved YOLOv2 object detection model. In 2018 6th International Conference on Multimedia Computing and Systems (ICMCS). IEEE. https://doi.org/10.1109/ICMCS.2018.8525895
2. Kim, K.-J., Kim, P.-K., Chung, Y.-S., & Choi, D.-H. (2018). Performance enhancement of YOLOv3 by adding prediction layers with spatial pyramid pooling for vehicle detection. In 2018 15th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS). IEEE. https://doi.org/10.1109/AVSS.2018.8639438
3. Mermertaş, F., Kapudere, M., Yerdoğan, M., Yenigün, B. (2022). BA-SA-RA Method-Based Educational Application for Special Students. Journal of Artificial Intelligence with Applications, 3(1), 20-23
4. Piştav Akmeşe, P. (2016). Türk işaret dili (TİD) eğitimi ve yüksek öğretim programlarında işaret dili dersi. Ahi Evran Üniversitesi Kırşehir Eğitim Fakültesi Dergisi, 17(3), 341–361.
5. Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) https://www.cv-foundation.org/openaccess/content_cvpr_2016/html/Redmon_You_Only_Look_CVPR_2016_paper.html
6. Roboflow. (2025). Türk İşaret Dili. Roboflow Universe. https://universe.roboflow.com/proje-qtjgs/turk-isaret-dili
7. Ultralytics. (n.d.). Ultralytics documentation. Retrieved from https://docs.ultralytics.com/
8. Türkiye İstatistik Kurumu. (2010), https://tuik.gov.tr
9. Türkiye Büyük Millet Meclisi. (2024).(Dönem 28, Yasama Yılı2). https://cdn.tbmm.gov.tr/KKBSPublicFile/D28/Y2/T7/WebOnergeMetni/ad7e3eb3-09be-4fc7-ba3a-affee920ca20.pdf
10. Tian, Y., Ye, Q., & Doermann, D. (2025). YOLOv12: Attention-centric real-time object detectors. arXiv. https://arxiv.org/abs/2502.12524
11. World Health Organization. (2024, 1 Mart). Deafness and hearing loss. https://www.who.int/news-room/fact-sheets/detail/deafness-and-hearing-loss
12. Wang, C.-Y., Yeh, I.-H., & Liao, H.-Y. M. (2024). YOLOv9: Learning what you want to learn using programmable gradient information. In Computer Vision – ECCV 2024 (pp. 1–21). Springer. https://doi.org/10.1007/978-3-031-72751-1_1
13. Wang, A., Chen, H., Liu, L., Chen, K., Lin, Z., Han, J., & Ding, G. (2024). YOLOv10: Real-time end-to-end object detection. In Advances in Neural Information Processing Systems, 37 (NeurIPS 2024). https://proceedings.neurips.cc/paper_files/paper/2024/hash/c34ddd05eb089991f06f3c5dc36836e0-Abstract-Conference.html
14. Yüksel, H. (1994). Bireylerarası İletişime Giriş. Eskişehir: Anadolu Üniversitesi Yayınları
15. Zhang, Y., Sun, Y., Wang, Z., & Jiang, Y. (2023). YOLOv7-RAR for urban vehicle detection. Sensors, 23(4), 1801. https://doi.org/10.3390/s23041801