An RGB-D Based Approach for Human Pose Estimation

With depth information more easily accessible even on mobile devices, leveraging RGB and depth information for RGB-D training provides a new way to enhance human pose estimation performance. In this paper, we propose an RGB-D based approach for human pose estimation. The main contributions of this paper are: 1) improving the accuracy and robustness of the model by utilizing depth image, 2) establishing a lightweight network architecture to improve the performance in detection speed, which makes it suitable for deployment on mobile devices. Qualitative and quantitative analyses on experimental results demonstrate that our model outperforms Open-Pose by 34% in detection speed, reduces model size to 42% at the same time. Our model also provides some advantages in specific background environments.