Skip to content

Latest commit

 

History

History
57 lines (40 loc) · 3.17 KB

README.md

File metadata and controls

57 lines (40 loc) · 3.17 KB

ArSL21L: Arabic sign language letter dataset benchmarking and an educational avatar for metaverse applications

Authors:

Ganzorig Batnasan, Munkhjargal Gochoo, Munkh-Erdene Otgonbold, Fady Alnajjar, Timothy K. Shih

samples_long

We present our collected and annotated Arabic Sign Language Letters Dataset (ArSL21L) consisting of 14202 images of 32 letter signs with various backgrounds collected from 50 people. We benchmarked our ArSL21L dataset on state-of-the-art object detection models, i.e., 4 versions of YOLOv5. Among the models, YOLOv5l achieved the best result with mAP of 0.83. Moreover, we provide comparison results of classification task between ArSL2018 dataset, the only Arabic sign language letter dataset for classification task, and our dataset by running classification task on in-house short video. The results revealed that the model trained on our dataset has a superior performance over the model trained on ArSL2018.

Dataset

Paper

Experimental results

signLanguage.mp4

The result of YOLOv5 models on ArSL21L dataset

Models Precision Recall mAP0.5 mAP.5:.95
YOLOv5s 0.953 0.9408 0.9784 0.7661
YOLOv5m 0.968 0.9468 0.9842 0.7768
YOLOv5l 0.9787 0.9766 0.9909 0.8306
YOLOv5x 0.9758 0.9743 0.9896 0.8224

Classification model results on Arabic datasets

ArSL2018 ArSL21L (Ours) Video
ResNext1 (ArSL2018) 99.2% 35% 69%
ResNext2 (ArSL21L(Ours)) 45.47% 91.04% 92%
CNN (ArSL2018) 92%

ddd

Citation

@INPROCEEDINGS{gbatnasan2021,
  author={Batnasan, Ganzorig and Gochoo, Munkhjargal and Otgonbold, Munkh-Erdene and Alnajjar, Fady and Shih, Timothy K},
  booktitle={2022 IEEE Global Engineering Education Conference (EDUCON)}, 
  title={ArSL21L: Arabic Sign Language Letter Dataset Benchmarking and an Educational Avatar for Metaverse Applications}, 
  year={2022},
  volume={},
  number={},
  pages={1814-1821},
  doi={10.1109/EDUCON52537.2022.9766497}}