Learning to Refine Human Pose Estimation

Fieraru, Mihai; Khoreva, Anna; Pishchulin, Leonid; Schiele, Bernt

doi:10.1109/cvprw.2018.00058

Cited by 109 publications

(61 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As shown in Table 6: AP comparison between PoseRefiner [10] and PoseFix on the PoseTrack 2018 validation set. The number in the parenthesis denotes the AP change from the input pose (i.e., Simple).…”

Section: Comparison With State-of-the-art Methodsmentioning

confidence: 99%

“…It is noticeable that our method can achieve better performance when a new state-of-the-art method is proposed by using it as the input pose of our method. We also compare the performance of the PoseFix with PoseRefiner [10] which has a similar approach to ours in Table 6. Table shows our PoseFix improve input pose significantly more than PoseRefiner [10].…”

Section: Comparison With State-of-the-art Methodsmentioning

confidence: 99%

“…We also compare the performance of the PoseFix with PoseRefiner [10] which has a similar approach to ours in Table 6. Table shows our PoseFix improve input pose significantly more than PoseRefiner [10].…”

Section: Comparison With State-of-the-art Methodsmentioning

confidence: 99%

See 2 more Smart Citations

PoseFix: Model-Agnostic General Human Pose Refinement Network

Moon

Chang

Lee

2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

162

View full text Add to dashboard Cite

Multi-person pose estimation from a 2D image is an essential technique for human behavior understanding. In this paper, we propose a human pose refinement network that estimates a refined pose from a tuple of an input image and input pose. The pose refinement was performed mainly through an end-to-end trainable multi-stage architecture in previous methods. However, they are highly dependent on pose estimation models and require careful model design. By contrast, we propose a model-agnostic pose refinement method. According to a recent study, state-of-the-art 2D human pose estimation methods have similar error distributions. We use this error statistics as prior information to generate synthetic poses and use the synthesized poses to train our model. In the testing stage, pose estimation results of any other methods can be input to the proposed method. Moreover, the proposed model does not require code or knowledge about other methods, which allows it to be easily used in the post-processing step. We show that the proposed approach achieves better performance than the conventional multi-stage refinement models and consistently improves the performance of various state-of-the-art pose estimation methods on the commonly used benchmark. The code is available in this https URL 1 .

show abstract

Section: Comparison With State-of-the-art Methodsmentioning

confidence: 99%

Section: Comparison With State-of-the-art Methodsmentioning

confidence: 99%

See 1 more Smart Citation

PoseFix: Model-Agnostic General Human Pose Refinement Network

Moon

Chang

Lee

2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

162

View full text Add to dashboard Cite

show abstract

“…The top-down approaches such as CPN 8) and AlphaPose 16) first detect person areas with a bounding box using, for example, faster R-CNN 12) in AlphaPose, then estimate the human pose in each de-tected person area. These approaches are highly sensitive to the accuracy of the person detector and are known to have difficulty in estimating poses of persons where there is occlusion 17) . The bottom-up approaches such as OpenPose 9) first predict the heatmap of all body joints, then connect the body joints for each person.…”

Section: Pose Estimationmentioning

confidence: 99%

[Papers] Keep Your Eye on the Ball: Detection of Kicking Motions in Multi-view 4K Soccer Videos

Tasaka

2020

MTA

View full text Add to dashboard Cite

For automated game analysis, it is essential to detect the kicking motions of players in soccer videos in order to understand each player's actions. This paper presents a fast and accurate approach to detecting kicking motions with a ball-centric window in multi-view 4K soccer videos. Based on powerful object detection techniques like SSD or YOLOv3 and pose estimation techniques like OpenPose or CPN, we propose novel solutions to overcome two challenges in 4K soccer videos. The first challenge is that it is basically too computationally heavy to process the massive amount of data in multi-view 4K videos. The solution to this challenge is that we only process a small portion (i.e. a ball-centric window) of 4K video, benefiting from an object tracking technique and homography transformation. The second challenge is that kicking motions may be incorrectly detected due to two factors. One is the absence of depth information and the other is the inaccuracy of pose estimation. We fuse multiple views to avoid the depth problem. In addition, we propose enlarging the person areas to effectively improve the accuracy of pose estimation. The experiments on real data from the J1 League demonstrate that the proposed approach achieves both faster and more accurate detection of kicking motions than conventional methods.

show abstract

“…However, even well pre-trained models may fail to accurately compute human poses in certain situations, e.g., when there exist occlusions and motion blurs. Even though there are auto matic pose refinement methods that correct the initial human pose estimation results [10,21], the refined human poses still contain errors that require further corrections. To address this problem, we provide a simple interface for modifying raw human pose estimation results.…”

Section: Session 11a: Authoring Animationmentioning

confidence: 99%