Automatic athlete pose estimation from images has recently received considerable attention from the computer vision community to understand the correct pose of athletes during training or competitions. However, human pose estimation from images or videos is still a challenging task because of inadequate training data, depth obscurity, and occlusion. In this paper, we presented a real-time 2D athlete pose estimation system based on OpenPose [1]. The system captured 2D positions of a hurdles athlete's body parts such as wrists, ankles, hips and knees from an image. For experiments, we recorded videos from a top-tier athlete in hurdling to infer motion timing and kinematic parameters. We also explore the performance of different feature extractorsin OpenPose in terms of accuracy and run-time. Experimental results show that the system with the extended version of VGG19 [2] as feature extractor outperforms the other algorithms.