Research on Human Pose Estimation
Ruiyang Sun, Huixin Yang, Lanfei Zhao, Yubin Sun · 2022 IEEE 5th International Conference on Electronics Technology (ICET) · 2022
We propose a convolutional neural network to estimate human pose. Firstly, a lightweight image scaling network is designed to scale the input image to a standard size image. This network would effectively suppress image feature loss caused by traditional interpolation method. Secondly, non-local convolution operator is introduced into the convolutional neural network, thereby increasing the receptive field. Finally, the resolution attention mechanism is introduced in the process of multiresolution feature fusion to improve the scale invariance of the network. Experimental results show that the scale invariance and accuracy for the proposed network is improved effectively.