Neural Network Image Segmentation for Sign Language Interpretation

Abstract

The use of neural networks to recognize and classify objects in images is a popular field in computer science. It is highly likely that an object in an image chosen for classification will have a representation matrix with significantly less pixels than the background or other elements of the image. As a result, the initial plan would be to divide or segment that object from the other portions of the image that are not essential for categorization. This also serves as the study's objective, for which we employ segmentation to separate the components essential to the classification procedure and assess any room for improvement in the final classification outcome. Mask Region Convolutional Neural Network was the model used for segmentation, and Convolutional Neural Network was the model used for classification. The study's findings demonstrate a notable improvement in the classification in the case of sign language. Further advancement of image segmentation models implies better more accurate results for classification models once they are combined. Keywords— Neural network, Image segmentation, Sign language, Classification, Mask Regional Convolutional Neural Network.