American Sign Language Classification

[Wikipedia] American Sign Language (ASL) is a natural language that serves as the predominant sign language of Deaf communities in the United States and most of Anglophone Canada. Besides North America, dialects of ASL and ASL-based creoles are used in many countries around the world, including much of West Africa and parts of Southeast Asia. ASL is also widely learned as a second language, serving as a lingua franca.

Data

To Download the ASL dataset -

• Click on the link given - https://www.kaggle.com/grassknoted/asl-alphabet
• Click on the link given - https://facundoq.github.io/unlp/sign_language_datasets/index.html

Getting Started

We will be using Python and Tensorflow to build model to classify signs ASL. The tensorflow graph of the architecture can be shown as:

CNN Architecture

• Input
• ⇊
• Convolution layer: 4x4 kernel , 8 filters
• Maxpooling: pool size – 8x8
• Convolution layer: 2x2 kernel , 16 filters
• Maxpooling: pool size – 4x4
• Flatten
• Fully Connected

Authors

• Mrityunjay Tripathi

License

This project is licensed under the Apache 2.0 License - see the License file for details.