Hand Gesture Recognition or Segmentation Using a Convolutional Neural Network with Synthetic Data Augmentation for Classification

Abstract

Recognition of hand gestures is a crucial component of human connection and communication. Hand gesture recognition can be used to overcome linguistic barriers as well as facilitate user engagement in the human-computer interface (HCI). The scientific community has recently given hand gesture recognition a lot of attention for a variety of applications, such as enhanced driver assistance systems, prosthetics, and robot controls. For example, hand gesture recognition may be used to understand sign language. In this paper, we proposed two models of convolutional neural network such as InceptionV3 and VGG16 for hand gesture recognition with synthetic data augmentation for classification using own build dataset. We will have to put in a lot of work to achieve the goal, including data preparation and cleaning. After pre-processing the dataset, we augment all of the data classification. After augmenting the data for the feed model, we employ a number of classification techniques to identify the pictures and ensure excellent accuracy in the outcome decision. Among them, Inception V3 and VGG-16 are the best outcomes. The presented architectures' classification results exhibit a high degree of accuracy, approaching 99.59%, and VGG16 has an accuracy of 97.21%.