The Performance of Thai Sign Language Recognition with 2D Convolutional Neural Network Based on NVIDIA Jetson Nano Developer Kit

Abstract

Thai Sign Language Recognition is a Thai Sign Language learning computer recognition. The system constructs an architecture of T-SLR by TSR-2DCNN based on NVIDIA Jetson Nano Developer Kit. It is a novelty of automatic translation TSL innovation and reveals the performance of feature extraction and classification to reduce crashed system, overloaded or automatic reboot while complicated processing occurs. The dataset contains 7 gestures in TSL, training images are 7,000 images and validation images are 700 images. The result compares with many techniques as shown that TSR-2DCNN can increase the performance of T-SLR in real-time, effectiveness with an accuracy of 0.9914 and loss of 0.03537.