American Sign Language Translator

• 

  Accuracy for the training model

• 

  Loss for the training model

Inspiration

American Sign Language has always been interesting to me. Although it's most commonly used by hard of hearing people, I believe it should be an essential method of communication. To begin this process, I wanted to develop an ASL translator to understand my friends who are fluent in ASL.

What it does

This program is supposed to take a video input and translate sign language.

How we built it

I developed a convoluted neural network (CNN) to understand the ASL alphabet dataset.

Challenges we ran into

As this was my first machine learning project, it took a lot of research and failures. Most of the time was spent learning about developing neural networks. I also spent a fair amount of time troubleshooting the kernels not working the way I intended.

Accomplishments that we're proud of

Although the project is still in progress due to a lack of video processing, I was able to train a model that has an accuracy of 99.63%.

What we learned

I learned how neural networks are developed and how to make them work efficiently.

What's next for American Sign Language Translator

In the future, I want to refine the model and make the video isolate the hand so that background noise does not affect the predicted labels. I would also like to incorporate the entire ASL library.

Built With

• jupyter
• python