Bristy Chakraborty; Masudur Rahman; Apurba Adhikary; Minoru W. Yoshida

Advancing Bangla Sign Language Detection through Dataset Creation, Model Comparison, and Deploy- ment

PDF (2361KB), PP.11-29

Views: 0 Downloads: 0

Author(s)

Bristy Chakraborty ¹ Masudur Rahman ^1,* Apurba Adhikary ¹ Minoru W. Yoshida ²

1. Department of Information and Communication Engineering, Noakhali Science and Technology University, Bangladesh

2. Department of Information System Creation, Kanagawa University, Japan

* Corresponding author.

DOI: https://doi.org/10.5815/ijem.2026.03.02

Received: 12 Jan. 2026 / Revised: 4 Feb. 2026 / Accepted: 14 Mar. 2026 / Published: 8 Jun. 2026

Index Terms

Bangla Sign Language, YOLOv8, BanglaSignSet, Deep Learning, Object Detection, Web Application

Abstract

Bangla Sign Language is a unique sign language. Due to a lack of interpreters, the hearing- and speech- impaired community face challenges while communicating with the broader community. Recent studies have been con- ducted to reduce the gap between these two communities. But most of the researchers used a dataset with a controlled environment. We know the performance of a system highly depends on dataset quality. In this paper, we have created a new dataset, “BanglaSignSet” including 46 unique signs with over 10k images. We have carefully annotated and labeled the images using Roboflow. Our proposed dataset, “BanglaSignSet” consists of images with high resolution, good qual- ity, and adequate variation in environment and person. The constructed dataset has been trained using the most recent deep learning model, such as YOLOv8. We have also implemented different versions of the YOLOv8 model, such as YOLOv8n, YOLOv8s, and YOLOv8m. Additionally, we evaluated EfficientNet-B0 as a classification-based baseline to broaden the experimental comparison. The performance of models has been measured using different evaluation metrics such as mAP, precision, recall, and f1 score. A comparative analysis has been conducted based on the performance of the model. By comparative analysis we found a well-suited model, YOLOv8n, to deploy into a web-based application. To find the suitable model to deploy, we have considered factors such as memory requirement and inference speed. We have integrated the YOLOv8n model into a web application using the Python language. We have also tested the web application on Android devices and laptops. The web application detects signs from image input successfully.

Cite This Paper

Bristy Chakraborty, Masudur Rahman, Apurba Adhikary, Minoru W. Yoshida, "Advancing Bangla Sign Language Detection through Dataset Creation, Model Comparison, and Deploy- ment", International Journal of Engineering and Manufacturing (IJEM), Vol.16, No.3, pp.11-29, 2026. DOI:10.5815/ijem.2026.03.02

Reference

[1]Addressing the rising prevalence of hearing loss. World Health Organization, 2018.url: https://www.who.int/publications/i/item/addressing-the-rising-prevalence-of-hearing-loss.
[2]James Garcia. How many people know sign language? Sign Station, 2021. url: https://signstation.org/how-many- people-know-sign-language/.
[3]Akhtar N Tarafder KH. Disabling impairment in the bangladeshi population. The Journal of Laryngology and otology, 2015. doi: 10.1017/s002221511400348x.
[4]Ying Xie Kshitij Bantupalli. American sign language recognition using deep learning and computer vision. 2018 IEEE International Conference on Big Data (Big Data), 2018. doi: 10.1109/BigData.2018.8622141.
[5]Sigberto Alarcon Viesca Brandon Garcia. Real-time american sign language recognition with convolutional neural networks. Neural Networks for Visual Recognition, 2016. doi: 10.13140/RG.2.2.31797.24805.
[6]Kam K.H. Ng C.K.M. Lee. American sign language recognition and training method with recurrent neural network. Expert Systems with Applications, 2021. doi: 10.1016/j.eswa.2020.114403.
[7]Kosin Chamnongthai Sunusi Bala Abdullahi. Spatial–temporal feature-based end-to-end fourier network for 3d sign language recognition. Expert Systems with Applications, 2024. doi: 10.1016/j.eswa.2024.123258.
[8]Saroj Kr. Biswas Soumen Das. Occlusion robust sign language recognition system for indian sign language using cnn and pose features. Multimedia Tools and Applications, 2024. doi: 10.1007/s11042-024-19068-0.
[9]Parteek Kumar Ankita Wadhawan. Sign language recognition systems: A decade systematic literature review. Archives of Computational Methods in Engineering, 2019. doi: 10.1007/s11831-019-09384-2.
[10]Varsha Singh Shagun Katoch. Indian sign language recognition system using surf with svm and cnn. Array, 2022. DOI: 10.1016/j.array.2022.100141.
[11]Nikita Sharma Ashish Sharma. Benchmarking deep neural network approaches for indian sign language recognition. Neural Computing and Applications, 2020. doi: 10.1007/s00521-020-05448-8.
[12]Arun Kumar Rinki Gupta. Indian sign language recognition using wearable sensors and multi-label classification. Computers & Electrical Engineering, 2021. doi: 10.1016/j.compeleceng.2020.106898.
[13]Ishak Pacal Melek Alaftekin. Real-time sign language recognition based on yolo algorithm. Neural Computing and Applications, 2024. doi: 10.1007/s00521-024-09503-6.
[14]Barıs¸ Bilen S, eyma Takır. Sentiment analysis in turkish sign language through facial expressions and hand gestures. IEEE Signal Processing and Communications Applications (SIU), 2024. doi: 10.1109/SIU61531.2024.10601084.
[15]Ipek ERDOGAN INCI MELIHA BAYTAS. Signer-independent sign language recognition with feature disentan- glement. Turkish Journal of Electrical Engineering and Computer Sciences, 2024. doi: 10.55730/1300-0632.4078.
[16]Md. Johir Raihan and Mainul Islam Labib. Bengali-sign: A machine learning-based bengali sign language interpre- tation for deaf and non-verbal people. Sensors, 2024. doi: 10.3390/s24165351.
[17]Md. Samir Imtiaz Sunanda Das. A hybrid approach for bangla sign language recognition using deep transfer learning model with random forest classifier. ScienceDirect, 2023. doi: 10.1016/j.eswa.2022.118914.
[18]A. S. M. Miah. Bensignnet: Bengali sign language alphabet recognition using concatenated segmentation and convolutional neural network. Applied Sciences, 2022. doi: 10.3390/app12083933.
[19]Md. Rasel Uddin Md. Monirul Islam. Bdslw-11: Dataset of bangladeshi sign language words for recognizing 11 daily useful bdsl words. Data in Brief, 2022. doi: 10.1016/j.dib.2022.108747.
[20]M. S Islam. Recognition bangla sign language using convolutional neural network. 2019 International Con- ference on Innovation and Intelligence for Informatics, Computing, and Technologies (3ICT), 2019. doi: 10.1109/3ICT.2019.8910301.
[21]Azmain Yakin Srizon Md. Mehedi Hasan. Classification of bengali sign language characters by applying a novel deep convolutional neural network. 2020 IEEE Region 10 Symposium (TENSYMP), 2020. doi: 10.1109/TEN- SYMP50017.2020.9230658.
[22]Sadia Sultana Sharmin Mousumi Md. Sanzidul Islam. Ishara-lipi: The first complete multi purpose open access dataset of isolated characters for bangla. International Conference on Bangla Speech and Language Process- ing(ICBSLP), 2018. doi: 10.1109/ICBSLP.2018.8554466.
[23]Jubair Hoque, O. B. Real time bangladeshi sign language detection using faster r-cnn. International Conference on Innovation in Engineering and Technology (ICIET), 2018. doi: 10.1109/CIET.2018.8660780.
[24]Nowshin Nawal Abdul Muntakim Rafi. Image-based bengali sign language alphabet recognition for deaf and dumb community. 2019 IEEE Global Humanitarian Technology Conference (GHTC), 2019. doi: 10.1109/GHTC46095.2019.9033031.
[25]Ultralytics.Explore ultralytics yolov8. Ultralytics YOLO Docs, 2025. https://docs.ultralytics.com/models/yolov8/.

International Journal of Engineering and Manufacturing (IJEM)