TRAINING OF A CONVOLUTIONAL NEURAL NETWORK FOR HAND GESTURE RECOGNIZING ON THE KAGGLE ASL ALPHABET DATASET

Abstract

Nowadays, hand gesture recognizing is important topic. It is used in virtual assistant work, for sign language translation, in virtual and augmented reality applications, and in entertainment services. The paper deals with the convolutional neural network training using different technologies. The neural network is trained to classify American manual alphabet and 3 extended gestures using photographs. The open access dataset Kaggle ASL Alphabet was used for training. Kaggle ASL Alphabet provides 87000 images of 29 classes for image classification and hand gesture recognizing.