Estimation of the Hand Pose using CNN in a Hand Pose Rally System
Haruto Morita, Akira Suganuma · 2024
A stamp rally is a project to collect stamps in a certain theme. We have been developing a Hand Pose Rally system that uses hand poses instead of stamps. In this rally system, it is essential to correctly recognize the hand pose. In this study, we used deep learning to develop the posture estimation part of the hand pose rally system. The learning has been performed with 32 different hand postures of finger bending and stretching patterns. The estimation accuracy of our method has been 95.7%.