Fig. 1. Overall structure of global hand pose estimation network based on pixel voting

Fig. 2. ASPP module structure

Fig. 3. Generate pose label map

Fig. 4. Aquisition of global hand pose

Fig. 5. Schematic diagram of hand 3D rotation axis

Fig. 6. Basic hand motion template

Fig. 7. Some examples of synthetic datasets

Fig. 8. Prediction error of global hand pose under different balance factors

Fig. 9. Global hand pose prediction accuracy of different methods on test dataset

Fig. 10. Global hand pose prediction results of depth image of real hand