Using VGG Models with Intermediate Layer Feature Maps for Static Hand Gesture Recognition

A hand gesture recognition system provides a robust and innovative solution to nonverbal communication through human–computer interaction. Deep learning models have excellent potential for usage in applications. To overcome related issues, most previous studies proposed new model architectures or fine-tuned pre-trained models. Furthermore, these relied on one standard dataset both training testing. Thus, the accuracy of is reasonable. Unlike works, current study investigates two deep with intermediate layers recognize static images. Both were tested different datasets, adjusted suit dataset, then trained under methods. First, initialized random weights from scratch. Afterward, examined as feature extractors. Finally, layers. Fine-tuning was conducted three levels: fifth, fourth, third blocks, respectively. The evaluated experiments using images Arabic sign language acquired conditions. This also image used experiments, plus other datasets. experimental results indicated that can be analysis showed fine-tuning fifth fourth blocks achieved best results. In particular, testing accuracies datasets 96.51%, 72.65%, 55.62% when block 96.50%, 67.03%, 61.09% first model. second approximately similar