In current times, an overwhelming number of world populations are special in that they do not have a large language due to the absence of their hearing capacity. People with hearing disability have their own language called Sign Language, but it is difficult for general individuals to understand[1]. Additionally, sign digits are a significant piece of communication dependent on gestures. So it is necessary for a machine interpreter to enable them to talk to general individuals. These days, it is noteworthy to make their language justifiable for the computer vision-based arrangements of the general person. We plan to build a model based on CNN to deal with the identification of Sign Language digits in this job exploration. To train (70 percent), validate (20 percent) and evaluate (10 percent) of the network, a dataset of 10 classes is used. To train and measure the accuracy of sign digits, we consider three distinct CNN network models. Among the three models, pre-trained CNN dependent transfer learning performs better with test accuracy of 92%.
Please see the link :- https://www.journalajrcos.com/index.php/AJRCOS/article/view/30154
No comments:
Post a Comment