Has My Algorithm Succeeded? An Evaluator for Human Pose Estimators

Jammalamadaka, Nataraj; Zisserman, Andrew; Eichner, Marcin; Ferrari, Vittorio; Jawahar, C. V.

doi:10.1007/978-3-642-33712-3_9

Cited by 31 publications

(32 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Inspired by the recent work of [19] we select the best component with the minimal uncertainty in the marginal posterior distributions of the body parts. The criteria used to measure the uncertainty is given by s(k, I) = ∑ …”

Section: Minimum Variance (Min-var)mentioning

confidence: 99%

Multi-view Pictorial Structures for 3D Human Pose Estimation

Amin¹,

Andriluka

Rohrbach

et al. 2013

Procedings of the British Machine Vision Conference 2013

137

154

View full text Add to dashboard Cite

Pictorial structure models are the de facto standard for 2D human pose estimation. Numerous refinements and improvements have been proposed such as discriminatively trained body part detectors, flexible body models, and local and global mixtures. While these techniques allow to achieve state-of-the-art performance for 2D pose estimation, they have not yet been extended to enable pose estimation in 3D. This paper thus proposes a multi-view pictorial structures model that builds on recent advances in 2D pose estimation and incorporates evidence across multiple viewpoints to allow for robust 3D pose estimation. We evaluate our multi-view pictorial structures approach on the HumanEva-I and MPII Cooking dataset. In comparison to related work for 3D pose estimation our approach achieves similar or better results while operating on single-frames only and not relying on activity specific motion models or tracking. Notably, our approach outperforms state-of-the-art for activities with more complex motions.

show abstract

Section: Minimum Variance (Min-var)mentioning

confidence: 99%

Multi-view Pictorial Structures for 3D Human Pose Estimation

Amin¹,

Andriluka

Rohrbach

et al. 2013

Procedings of the British Machine Vision Conference 2013

137

154

View full text Add to dashboard Cite

show abstract

“…We encode the uncertainty in pose estimation by computing the L2 norm of the covariance matrix corresponding to the strongest mode in the marginal posterior distribution of each part in each view. This is the same criteria as used for component selection in [1] and is similar to the features used in [11]. 3.…”

Section: Confident Examples Miningmentioning

confidence: 99%

“…Various approaches for estimating the confidence of the pose prediction have been considered in the literature, ranging from models that are discriminatively trained for both detection and pose estimation [18] to specialized methods that estimate confidence based on the combination of features as a post-processing step [11]. In this paper we follow the direction similar to [11] but also employ features based on the 3D pose reconstruction, which we find to be highly effective for filtering out incorrect pose estimates. Overall we make the following contributions.…”

Section: Introduction and Related Workmentioning

confidence: 99%

“…As a second contribution we evaluate two approaches to assess the accuracy of 3D pose estimation. The first is to train a discriminative model based on various pose quality features as in [11], and the second is to consider the agreement of an ensemble of several independently trained models on the 3D pose output. An interesting finding of our evaluation is that pose agreement alone performs on-par or better than the discriminatively trained confidence predictor.…”

Section: Introduction and Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Test-Time Adaptation for 3D Human Pose Estimation

Amin

Müller

Bulling

et al. 2014

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. In this paper we consider the task of articulated 3D human pose estimation in challenging scenes with dynamic background and multiple people. Initial progress on this task has been achieved building on discriminatively trained part-based models that deliver a set of 2D body pose candidates that are then subsequently refined by reasoning in 3D [1,4,5]. The performance of such methods is limited by the performance of the underlying 2D pose estimation approaches. In this paper we explore a way to boost the performance of 2D pose estimation based on the output of the 3D pose reconstruction process, thus closing the loop in the pose estimation pipeline. We build our approach around a component that is able to identify true positive pose estimation hypotheses with high confidence. We then either retrain 2D pose estimation models using such highly confident hypotheses as additional training examples, or we use similarity to these hypotheses as a cue for 2D pose estimation. We consider a number of features that can be used for assessing the confidence of the pose estimation results. The strongest feature in our comparison corresponds to the ensemble agreement on the 3D pose output. We evaluate our approach on two publicly available datasets improving over state of the art in each case.

show abstract

“…human-annotated data [3]- [5] or precision instrumentation [6]- [9]. These external sources of feedback which are not available during normal operation are known as "ground truth".…”

Section: Introductionmentioning

confidence: 99%

Introspective Evaluation of Perception Performance for Parameter Tuning without Ground Truth

Hu¹,

Kantor²

2017

Robotics: Science and Systems XIII

View full text Add to dashboard Cite

Abstract-Modern perception systems are notoriously complex, featuring dozens of interacting parameters that must be tuned to achieve good performance. Conventional tuning approaches require expensive ground truth, while heuristic methods are difficult to generalize. In this work, we propose an introspective ground-truth-free approach to evaluating the performance of a generic perception system. By using the posterior distribution estimate generated by a Bayesian estimator, we show that the expected performance can be estimated efficiently and without ground truth. Our simulated and physical experiments in a demonstrative indoor ground robot state estimation application show that our approach can order parameters similarly to using a ground-truth system, and is able to accurately identify top-performing parameters in varying contexts. In contrast, baseline approaches that reason only about observation log-likelihood fail in the face of challenging perceptual phenomena.

show abstract

Has My Algorithm Succeeded? An Evaluator for Human Pose Estimators

Cited by 31 publications

References 23 publications

Multi-view Pictorial Structures for 3D Human Pose Estimation

Multi-view Pictorial Structures for 3D Human Pose Estimation

Test-Time Adaptation for 3D Human Pose Estimation

Introspective Evaluation of Perception Performance for Parameter Tuning without Ground Truth

Contact Info

Product

Resources

About