Optimizing Machine Learning Pipelines: Scalable Mobile-Based Tamil Sign Language Recognition

Signs are used by humans to communicate, bridging the communication barrier among the general public and bridging the differences between the specially-abled. Human Translators became ineffective by the 21st Century prompting the introduction of Automatic Sign Language Recognition (ASLR) systems which recognize patterns based on a person’s gestures using Artificial Intelligence. While ASLR has made strides in global languages like English, it has been limited in its application to vernacular languages like Tamil. This project aims to create an ASLR App for the Tamil alphabet. Our approach towards this project is a Structured Evolution built up by analyzing and selecting techniques in each of the four conventional phases of machine learning. This derived pipeline of Machine Learning Techniques has been efficiently deployed into a web server and made available to access through a user-friendly web interface.