Research of Neural Network Models of Gesture Recognition in the Presence of Negative Factors

The problem of automatic recognition of gesture images for computer vision systems is considered. The process of preparing the initial data, creating a training and test dataset is described. A custom dataset has been created, as well as integration and preparation of external data. Research of popular neural network methods of sign language recognition has been conducted and an assessment of their effectiveness in the presence of negative factors has been obtained. Recommendations are given to improve the quality of gesture recognition.