Lightweight Human Pose Estimation with Attention Mechanism

With the development of computer vision technology, human pose estimation as an indispensable part of human-computer interaction. Although Light-Weight High-Resolution Network-30 has Lower number of parameters, the problem of in-sensitivity to local information and inaccurate prediction of keypoint locations. Some lightweight models lose in accuracy. To improve the accuracy of human pose estimation, this paper proposed a method which combines the Convolutional Block Attention Module attention mechanism with a lightweight high resolution network-30. The method proposed in this paper was evaluated on the human keypoint detection datasets COCO and MPII, and compared with the current mainstream methods. The experimental results show that the proposed method effectively improve the accuracy of the model while ensuring minimal loss in computation time.