mlsanityData to 14 Oct 2026

On the Calibration of Human Pose Estimation

arXiv 2311.17105 ↗First posted 28 Nov 2023cs.CVcited by 10

Not on the map. Why?

Abstract (arXiv)

Most 2D human pose estimation frameworks estimate keypoint confidence in an ad-hoc manner, using heuristics such as the maximum value of heatmaps. The confidence is part of the evaluation scheme, e.g., AP for the MSCOCO dataset, yet has been largely overlooked in the development of state-of-the-art methods. This paper takes the first steps in addressing miscalibration in pose estimation. From a calibration point of view, the confidence should be aligned with the pose accuracy. In practice, existing methods are poorly calibrated. We show, through theoretical analysis, why a miscalibration gap exists and how to narrow the gap. Simply predicting the instance size and adjusting the confidence function gives considerable AP improvements. Given the black-box nature of deep neural networks, however, it is not possible to fully close this gap with only closed-form adjustments. As such, we go one step further and learn network-specific adjustments by enforcing consistency between confidence and pose accuracy. Our proposed Calibrated ConfidenceNet (CCNet) is a light-weight post-hoc addition that improves AP by up to 1.4% on off-the-shelf pose estimation frameworks. Applied to the downstream task of mesh recovery, CCNet facilitates an additional 1.0mm decrease in 3D keypoint error.

Key facts. Cited by 10 papers in the map (data through 14 Oct 2026). Builds on 31 of its 40 references that are in the map. First posted on arXiv 28 Nov 2023.

Builds on

31 of 40 references.

References · 31

  1. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  2. 2023Adding Conditional Control to Text-to-Image Diffusion Modelscited by 5,503
  3. 2017On Calibration of Modern Neural Networkscited by 4,960
  4. 2016Simple and Scalable Predictive Uncertainty Estimation using Deep Ensemblescited by 3,926
  5. 2017What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?cited by 2,352
  6. 2019Deep High-Resolution Representation Learning for Human Pose Estimationcited by 1,339
  7. 2019Learning to Reconstruct 3D Human Pose and Shape via Model-fitting in the Loopcited by 627
  8. 2018Simple Baselines for Human Pose Estimation and Trackingcited by 502
  9. 2022ViTPose: Simple Vision Transformer Baselines for Human Pose Estimationcited by 456
  10. 2019Deep Evidential Regressioncited by 355
  11. 2013DeepPose: Human Pose Estimation via Deep Neural Networkscited by 305
  12. 2017Integral Human Pose Regressioncited by 256
19 more
  1. 2019Addressing Failure Prediction by Learning Model Confidencecited by 186
  2. 2020Whole-Body Human Pose Estimation in the Wildcited by 178
  3. 2019Distribution-Aware Coordinate Representation for Human Pose Estimationcited by 165
  4. 2018Uncertainty Estimates and Multi-Hypotheses Networks for Optical Flowcited by 135
  5. 2017Towards Accurate Multi-person Pose Estimation in the Wildcited by 131
  6. 2021Human Pose Regression with Residual Log-likelihood Estimationcited by 131
  7. 2021Pose Recognition with Cascade Transformerscited by 99
  8. 2019Distribution Calibration for Regressioncited by 89
  9. 2021Probabilistic Monocular 3D Human Pose Estimation with Normalizing Flowscited by 85
  10. 2020Point-Set Anchors for Object Detection, Instance Segmentation and Pose Estimationcited by 57
  11. 2017Benchmarking and Error Diagnosis in Multi-Instance Pose Estimationcited by 48
  12. 2022Poseur: Direct Human Pose Regression with Transformerscited by 43
  13. 2021Stochastic Optimization of Areas Under Precision-Recall Curves with Provable Convergencecited by 35
  14. 2021Calibrated and Sharp Uncertainties in Deep Learning via Density Estimationcited by 33
  15. 2022MUAD: Multiple Uncertainties for Autonomous Driving, a benchmark for multiple uncertainty types and taskscited by 24
  16. 2023Multiclass Confidence and Localization Calibration for Object Detectioncited by 16
  17. 2021SLURP: Side Learning Uncertainty for Regression Problemscited by 12
  18. 2023Discretization-Induced Dirichlet Posterior for Robust Uncertainty Quantification on Regressioncited by 7
  19. 2022Multi-hypothesis 3D human pose estimation metrics favor miscalibrated distributionscited by 6

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 5; 2025: 3; 2026: 2.

10 papers.

  1. 2024ProbPose: A Probabilistic Approach to 2D Human Pose Estimationcited by 9
  2. 2024KITRO: Refining Human Mesh by 2D Clues and Kinematic-tree Rotationcited by 5
  3. 2024Utilizing Uncertainty in 2D Pose Detectors for Probabilistic 3D Human Mesh Recoverycited by 4
  4. 2024Platypose: Calibrated Zero-Shot Multi-Hypothesis 3D Human Motion Estimationcited by 3
  5. 2024Detection, Pose Estimation and Segmentation for Multiple Bodies: Closing the Virtuous Circlecited by 2
  6. 2025Continuous Normalizing Flows for Uncertainty-Aware Human Pose Estimationcited by 2
  7. 2026Flow Matching for Probabilistic Monocular 3D Human Pose Estimationcited by 2
  8. 2025Semantics-aware Test-time Adaptation for 3D Human Pose Estimationcited by 1
  9. 2025Online Test-time Adaptation for 3D Human Pose Estimation: A Practical Perspective with Estimated 2D Posescited by 1
  10. 2026MEOM: Multi-View Expected-OKS Maximization for Human Pose Triangulationcited by 0