# An Empirical Study of the Collapsing Problem in Semi-Supervised 2D Human Pose Estimation

arXiv: https://arxiv.org/abs/2011.12498 · HTML: https://mlsanity.com/paper/2011.12498/

## Key facts

- Cited by 21 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 25 Nov 2020.

## Built on by (most cited first, 21 of 21)

- [Human Pose as Compositional Tokens](https://mlsanity.com/paper/2303.11638/) (2023)
- [A Unified Framework for Domain Adaptive Pose Estimation](https://mlsanity.com/paper/2204.00172/) (2022)
- [Efficient Image Pre-Training with Siamese Cropped Masked Autoencoders](https://mlsanity.com/paper/2403.17823/) (2024)
- [FisherMatch: Semi-Supervised Rotation Regression via Entropy-based Filtering](https://mlsanity.com/paper/2203.15765/) (2022)
- [Semi-Supervised 2D Human Pose Estimation Driven by Position Inconsistency Pseudo Label Correction Module](https://mlsanity.com/paper/2303.04346/) (2023)
- [Domain-Adaptive 2D Human Pose Estimation via Dual Teachers in Extremely Low-Light Conditions](https://mlsanity.com/paper/2407.15451/) (2024)
- [Learning Articulated Shape with Keypoint Pseudo-labels from Web Images](https://mlsanity.com/paper/2304.14396/) (2023)
- [Pose-MUM : Reinforcing Key Points Relationship for Semi-Supervised Human Pose Estimation](https://mlsanity.com/paper/2203.07837/) (2022)
- [Rethinking Self-training for Semi-supervised Landmark Detection: A Selection-free Approach](https://mlsanity.com/paper/2404.04556/) (2024)
- [From Sharp to Blur: Unsupervised Domain Adaptation for 2D Human Pose Estimation Under Extreme Motion Blur Using Event Cameras](https://mlsanity.com/paper/2507.22438/) (2025)
- [Active Semi-Supervised Learning by Exploring Per-Sample Uncertainty and Consistency](https://mlsanity.com/paper/2303.08978/) (2023)
- [Semi-Supervised Unconstrained Head Pose Estimation in the Wild](https://mlsanity.com/paper/2404.02544/) (2024)
- [Location-free Human Pose Estimation](https://mlsanity.com/paper/2205.12619/) (2022)
- [Generating Unbiased Pseudo-labels via a Theoretically Guaranteed Chebyshev Constraint to Unify Semi-supervised Classification and Regression](https://mlsanity.com/paper/2311.01782/) (2023)
- [Expanding mmWave Datasets for Human Pose Estimation with Unlabeled Data and LiDAR Datasets](https://mlsanity.com/paper/2603.14507/) (2026)
- [Denoising and Selecting Pseudo-Heatmaps for Semi-Supervised Human Pose Estimation](https://mlsanity.com/paper/2310.00099/) (2023)
- [Modeling the Uncertainty with Maximum Discrepant Students for Semi-supervised 2D Pose Estimation](https://mlsanity.com/paper/2311.01770/) (2023)
- [Boosting Semi-Supervised 2D Human Pose Estimation by Revisiting Data Augmentation and Consistency Training](https://mlsanity.com/paper/2402.11566/) (2024)
- [Semi-supervised 2D Human Pose Estimation via Adaptive Keypoint Masking](https://mlsanity.com/paper/2404.14835/) (2024)
- [Decorrelating Structure via Adapters Makes Ensemble Learning Practical for Semi-supervised Learning](https://mlsanity.com/paper/2408.04150/) (2024)
- [Uncertainty-aware Long-tailed Weights Model the Utility of Pseudo-labels for Semi-supervised Learning](https://mlsanity.com/paper/2503.09974/) (2025)

## Abstract (arXiv)

Semi-supervised learning aims to boost the accuracy of a model by exploring unlabeled images. The state-of-the-art methods are consistency-based which learn about unlabeled images by encouraging the model to give consistent predictions for images under different augmentations. However, when applied to pose estimation, the methods degenerate and predict every pixel in unlabeled images as background. This is because contradictory predictions are gradually pushed to the background class due to highly imbalanced class distribution. But this is not an issue in supervised learning because it has accurate labels. This inspires us to stabilize the training by obtaining reliable pseudo labels. Specifically, we learn two networks to mutually teach each other. In particular, for each image, we compose an easy-hard pair by applying different augmentations and feed them to both networks. The more reliable predictions on easy images in each network are used to teach the other network to learn about the corresponding hard images. The approach successfully avoids degeneration and achieves promising results on public datasets. The source code and pretrained models have been released at https://github.com/xierc/Semi_Human_Pose.