The dataset contains images of 10 different Bangladeshi sign letters. 40 different images of different skin color, hand measurements and rotational difference for each alphabet has been captured and processed to curate a suitable dataset of 400 images for the system. Later, to test the strength of our algorithm, the contrast of the samples has been varied and the low-quality images have been used as the test subject. To examine whether the system is rotationally invariant or not, up to 5◦ rotational change has been made and used to test the samples.

Please cite if you use this dataset for any research work:

@INPROCEEDINGS{9019931,
  author={F. {Sadik} and M. R. {Subah} and A. G. {Dastider} and S. A. {Moon} and S. S. {Ahbab} and S. A. {Fattah}},
  booktitle={2019 IEEE International WIE Conference on Electrical and Computer Engineering (WIECON-ECE)}, 
  title={Bangla Sign Language Recognition with Skin Segmentation and Binary Masking}, 
  year={2019},
  volume={},
  number={},
  pages={1-5},}