Multi-Scale Tiny Region Gesture Recognition Towards 3D Object Manipulation In Industrial Design
Ledan Qian, Xiao Yu Zhou, Xuankang Mou, Yi Li · 2021
This paper proposed a smart 3D virtual object manipulation by gesture recognition with deep network training for multiply scalar tiny region targets. We introduce the famous YOLOv5 based improved and enhanced target detection deep networks, and provide effective and high efficiency gesture recognition method towards 3D object manipulation. Meanwhile, combined with the adaptive anchor frame calculation and target recognition accelerator, our method can capture the tiny gesture region and for multi-scale of image detection. The method can be efficiently applied to the virtual control with different hand gestures. The experiment results are high efficiency and accuracy. It has high application value in the field of industrial design based on gesture manipulation.