Abstract: A 2D single-person pose estimation method based on improved lightweight hourglass model is proposed. The inverse residual convolution was used to construct an improved lightweight hourglass model, which effectively reduced the number of parameters and the amount of calculation. Multi-scale feature fusion was used to improve the key point detection ability of the lightweight model under occlusion. The introduction of the knowledge distillation method enabled the improved model to significantly reduce the computing resources required for training and deployment when the detection accuracy was slightly reduced. The detection results of the MPII data set and practical application show that the improved lightweight hourglass model can effectively detect the key points of human bones, with good real-time performance and strong robustness, and can overcome the occlusion problem to a certain extent.