AttV19Net: Attention Based VGG 19 Network for Hand Gesture Recognition with a Prototype

Deep learning algorithms, particularly convolutional neural networks (CNNs), have shown promise in hand gesture recognition. However, challenges such as high variance in predictions, overfitting, and errors with similar gestures persist. To address these issues, we introduce AttV19Net, a novel enhancement to the VGG 19 model incorporating a spatial attention module. This enhancement allows the model to more effectively emphasize crucial image details. Our proposed method was rigorously evaluated on two publicly available datasets, Massey and ISL, achieving outstanding accuracies of 94.44% and 97.36%, respectively, thus surpassing existing state-of-the-art methods. Additionally, we developed a user-friendly graphical interface (GUI) prototype to demonstrate the practical application of our approach.