Performance Analysis of Transfer Learning Techniques for Hand-Gesture Controlled Robot

Mobile robots are becoming increasingly integrated into many domains such as health care, manufacturing, and personal assistance that makes the demand of hand gesture-controlled robots that efficient in recognition are become surge in use. The current study presents the hand gesture recognition for human-robot interaction in three cases by transfer learning approach with pre-trained models such as VGG16, AlexNet, EfficientNetB4, InceptionV3 and standard machine learning techniques such as Random Forest, Logistic Regression and ensemble method such as CatBoost. The study is to compare the effectiveness of these methods to identify the optimal model for accurate and efficient hand-gesture recognition, improving the control mechanisms of robots in real-world applications. The InceptionV3 and EfficientNetB4 models performed well with an accuracy of around 95 % in most of the cases and ensemble method CatBoost performed well with an accuracy of 97-98 % in all cases.