SignNet: A Deep Learning Architecture for Accurate Sign Language Recognition from Images

Abstract

Sign Language, as the word is self-explanatory, it means using the signs, or herein we can also call them as gestures for inter-communication between people who are specially-abled and even with those who try to express themselves with gestures. Sign Language System is a treading road for the field of deep learning and communication. There are models made every year that surpass all others and succeed state of the art mark achieving best results. Convolutional neural networks have evolved a lot in past years providing unbelievable outcomes and results. Thus, using ASL dataset and state of the art models like VGG16 and ResNet50, making a SLR system via transfer learning as well as understanding in-depth and detailed convolutional neural network structure and steps involved in it. It is an irrefutable fact that there are problems when it comes to communicating with disabled people, thus research regarding sign languages could be a great help for disabled people as well as for deep learning as a whole.