Learning effective human pose estimation from inaccurate annotation

Johnson, Samuel; Everingham, Mark

doi:10.1109/cvpr.2011.5995318

Cited by 384 publications

(285 citation statements)

References 16 publications

Supporting

Mentioning

284

Contrasting

Order By: Relevance

“…We show that reducing the effect of occluded body parts can provide better appearance modeling to improve the performance in human pose estimation. For appearance modeling, our approach is more efficient than the previous work [25] that discards sample images including occluded body parts. We have shown that our method leads to superior results than the base methods [1], [11].…”

Section: Resultsmentioning

confidence: 92%

“…In our model we weight a sample containing an occluded body part in order to mitigate the adverse effect of occlusion. The adverse effect of occlusion is partly suppressed in appearance modeling proposed in [25]. In [25], training data including occluded body parts is not used even if other body parts are visible and can be used for appearance modeling.…”

Section: Related Workmentioning

confidence: 99%

“…Another approach regarding occlusion is proposed by Johnson and Everingham in [25]. They use the dataset which has (incomplete) occlusion annotations where occluded body parts are not annotated.…”

Section: Related Workmentioning

confidence: 99%

“…In [25], training data including occluded body parts is not used even if other body parts are visible and can be used for appearance modeling. The proposed approach is more efficient than [25] in a sense that we fully utilize all training data, which are made with high annotation cost.…”

Section: Related Workmentioning

confidence: 99%

See 3 more Smart Citations

Occluded Appearance Modeling with Sample Weighting for Human Pose Estimation

Kawana

Ukita

2017

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYThis paper proposes a method for human pose estimation in still images. The proposed method achieves occlusion-aware appearance modeling. Appearance modeling with less accurate appearance data is problematic because it adversely affects the entire training process. The proposed method evaluates the effectiveness of mitigating the influence of occluded body parts in training sample images. In order to improve occlusion evaluation by a discriminatively-trained model, occlusion images are synthesized and employed with non-occlusion images for discriminative modeling. The score of this discriminative model is used for weighting each sample in the training process. Experimental results demonstrate that our approach improves the performance of human pose estimation in contrast to base models. key words: human pose estimation, pictorial structure models, occlusion

show abstract

Section: Resultsmentioning

confidence: 92%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

Occluded Appearance Modeling with Sample Weighting for Human Pose Estimation

Kawana

Ukita

2017

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

show abstract

“…The most acknowledged model of this category is pictorial structures [4][5][6]. Currently, most of the state-of-the-art approaches for human pose estimation rely on pictorial structures [7,8,2,3]. Those approaches have delivered promising results on standard evaluation datasets, but they build on complex appearance and body prior models.…”

Section: Introductionmentioning

confidence: 99%

Holistic Human Pose Estimation with Regression Forests

Belagiannis

Amann

Navab

et al. 2014

Articulated Motion and Deformable Objects

View full text Add to dashboard Cite

Abstract. In this work, we address the problem of human pose estimation in still images by proposing a holistic model for learning the appearance of the human body from image patches. These patches, which are randomly chosen, are used for extracting features and training a regression forest. During training, a mapping between image features and human poses, defined by joint offsets, is learned; while during prediction, the body joints are estimated with an efficient mode-seeking algorithm. In comparison to other holistic approaches, we can recover body poses from occlusion or noisy data. We demonstrate the power of our method in two publicly available datasets and propose a third one. Finally, we achieve state-of-the-art results in comparison to other approaches.

show abstract