Multi-person 3D Pose Estimation From Monocular Image Sequences

With the rapid development of visual action recognition,3D human skeleton reconstruction in a single image and image sequences has attracted increasing attention in recent years.Compared with 2D human skeleton,3D human skeleton generally leads to better performance of action recognition,due to the rotation invariance of 3D human skeleton.As a result,3D pose estimation will greatly promote the development of action recogniton.However,to the best of our knowlegde,most existing work did 3D pose estimation on the single person in the controlled environmnt.So it is very useful and practical to solve the problem of multi-person 3D pose estimation.In this paper,we tackle the problem of multi-person 3D human pose estimation based on monocular image sequence in a three-step framework:(1)we detect 2D human skeletons in each frame across the image sequence;(2)we track each person through the image sequence and identify the sequence of 2D skeletons for each person;(3)we reconstruct the 3D human skeleton for each person from the detected 2D human joints,by using prelearned base poses and considering the temporal smoothness.The main contribution of the proposed method is that our framework can tackle the problem of multi-person 3D human pose estimation.We evaluate our framework on the Human3.6M dataset and the multi-person image sequence captured by ourselves.The results on the Human3.6M dataset demonstrate that the proposed method outperforms the current state-of-art methods.Besides,the estimated 3D human poses in multi-person image sequences show the advantage of our proposed framework in a qualitative fashion.