# Out-of-Domain Human Mesh Reconstruction via Dynamic Bilevel Online Adaptation

arXiv: https://arxiv.org/abs/2111.04017 · HTML: https://mlsanity.com/paper/2111.04017/

## Key facts

- Cited by 24 papers in the map (data through 14 Oct 2026).
- Builds on 66 of its 87 references that are in the map.
- First posted on arXiv 7 Nov 2021.

## Builds on (most cited first, 30 of 66)

- [Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks](https://mlsanity.com/paper/1703.03400/) (2017)
- [Expressive Body Capture: 3D Hands, Face, and Body from a Single Image](https://mlsanity.com/paper/1904.05866/) (2019)
- [Deeper, Broader and Artier Domain Generalization](https://mlsanity.com/paper/1710.03077/) (2017)
- [End-to-end Recovery of Human Shape and Pose](https://mlsanity.com/paper/1712.06584/) (2017)
- [Keep it SMPL: Automatic Estimation of 3D Human Pose and Shape from a Single Image](https://mlsanity.com/paper/1607.08128/) (2016)
- [Learning to Reconstruct 3D Human Pose and Shape via Model-fitting in the Loop](https://mlsanity.com/paper/1909.12828/) (2019)
- [Domain Generalization via Invariant Feature Representation](https://mlsanity.com/paper/1301.2115/) (2013)
- [VIBE: Video Inference for Human Body Pose and Shape Estimation](https://mlsanity.com/paper/1912.05656/) (2019)
- [Monocular 3D Human Pose Estimation In The Wild Using Improved CNN Supervision](https://mlsanity.com/paper/1611.09813/) (2016)
- [Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results](https://mlsanity.com/paper/1703.01780/) (2017)
- [Generalizing to Unseen Domains via Adversarial Data Augmentation](https://mlsanity.com/paper/1805.12018/) (2018)
- [End-to-End Human Pose and Mesh Reconstruction with Transformers](https://mlsanity.com/paper/2012.09760/) (2020)
- [Learning from Synthetic Humans](https://mlsanity.com/paper/1701.01370/) (2017)
- [PARE: Part Attention Regressor for 3D Human Body Estimation](https://mlsanity.com/paper/2104.08527/) (2021)
- [Learning 3D Human Dynamics from Video](https://mlsanity.com/paper/1812.01601/) (2018)
- [HybrIK: A Hybrid Analytical-Neural Inverse Kinematics Solution for 3D Human Pose and Shape Estimation](https://mlsanity.com/paper/2011.14672/) (2020)
- [I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image](https://mlsanity.com/paper/2008.03713/) (2020)
- [Convolutional Mesh Regression for Single-Image Human Shape Reconstruction](https://mlsanity.com/paper/1905.03244/) (2019)
- [Generalizing Across Domains via Cross-Gradient Training](https://mlsanity.com/paper/1804.10745/) (2018)
- [Mesh Graphormer](https://mlsanity.com/paper/2104.00272/) (2021)
- [Domain Generalization for Object Recognition with Multi-task Autoencoders](https://mlsanity.com/paper/1508.07680/) (2015)
- [PyMAF: 3D Human Pose and Shape Regression with Pyramidal Mesh Alignment Feedback Loop](https://mlsanity.com/paper/2103.16507/) (2021)
- [Episodic Training for Domain Generalization](https://mlsanity.com/paper/1902.00113/) (2019)
- [Learning to Estimate 3D Human Pose and Shape from a Single Color Image](https://mlsanity.com/paper/1805.04092/) (2018)
- [Monocular, One-stage, Regression of Multiple 3D People](https://mlsanity.com/paper/2008.12272/) (2020)
- [Unite the People: Closing the Loop Between 3D and 2D Human Representations](https://mlsanity.com/paper/1701.02468/) (2017)
- [Neural Body Fitting: Unifying Deep Learning and Model-Based Human Pose and Shape Estimation](https://mlsanity.com/paper/1808.05942/) (2018)
- [Populating 3D Scenes by Learning Human-Scene Interaction](https://mlsanity.com/paper/2012.11581/) (2020)
- [Probabilistic Modeling for Human Mesh Recovery](https://mlsanity.com/paper/2108.11944/) (2021)
- [Perceiving 3D Human-Object Spatial Arrangements from a Single Image in the Wild](https://mlsanity.com/paper/2007.15649/) (2020)

## Built on by (most cited first, 24 of 24)

- [CLIFF: Carrying Location Information in Full Frames into Human Pose and Shape Estimation](https://mlsanity.com/paper/2208.00571/) (2022)
- [EgoHumans: An Egocentric 3D Multi-Human Benchmark](https://mlsanity.com/paper/2305.16487/) (2023)
- [3D Human Mesh Estimation from Virtual Markers](https://mlsanity.com/paper/2303.11726/) (2023)
- [CIMI4D: A Large Multimodal Climbing Motion Dataset under Human-scene Interactions](https://mlsanity.com/paper/2303.17948/) (2023)
- [SportsPose -- A Dynamic 3D sports pose dataset](https://mlsanity.com/paper/2304.01865/) (2023)
- [Cyclic Test-Time Adaptation on Monocular Video for 3D Human Mesh Reconstruction](https://mlsanity.com/paper/2308.06554/) (2023)
- [Deep learning for 3D human pose estimation and mesh recovery: A survey](https://mlsanity.com/paper/2402.18844/) (2024)
- [KITRO: Refining Human Mesh by 2D Clues and Kinematic-tree Rotation](https://mlsanity.com/paper/2405.19833/) (2024)
- [MoCap-to-Visual Domain Adaptation for Efficient Human Mesh Estimation from 2D Keypoints](https://mlsanity.com/paper/2404.07094/) (2024)
- [BlanketSet -- A clinical real-world in-bed action recognition and qualitative semi-synchronised MoCap dataset](https://mlsanity.com/paper/2210.03600/) (2022)
- [A Survey of Advanced Computer Vision Techniques for Sports](https://mlsanity.com/paper/2301.07583/) (2023)
- [STRIDE: Single-video based Temporally Continuous Occlusion-Robust 3D Pose Estimation](https://mlsanity.com/paper/2312.16221/) (2023)
- [Uncertainty-Aware Testing-Time Optimization for 3D Human Pose Estimation](https://mlsanity.com/paper/2402.02339/) (2024)
- [BlanketGen - A synthetic blanket occlusion augmentation pipeline for MoCap datasets](https://mlsanity.com/paper/2210.12035/) (2022)
- [G3R: Generating Rich and Fine-grained mmWave Radar Data from 2D Videos for Generalized Gesture Recognition](https://mlsanity.com/paper/2404.14934/) (2024)
- [ESCAPE: Energy-based Selective Adaptive Correction for Out-of-distribution 3D Human Pose Estimation](https://mlsanity.com/paper/2407.14605/) (2024)
- [Semantics-aware Test-time Adaptation for 3D Human Pose Estimation](https://mlsanity.com/paper/2502.10724/) (2025)
- [Online Test-time Adaptation for 3D Human Pose Estimation: A Practical Perspective with Estimated 2D Poses](https://mlsanity.com/paper/2503.11194/) (2025)
- [MobileOcc: A Human-Aware Semantic Occupancy Dataset for Mobile Robots](https://mlsanity.com/paper/2511.16949/) (2025)
- [Lifelong Domain Adaptive 3D Human Pose Estimation](https://mlsanity.com/paper/2512.23860/) (2025)
- [PRIMA: Boosting Animal Mesh Recovery with Biological Priors and Test-Time Adaptation](https://mlsanity.com/paper/2606.02366/) (2026)
- [3DHR-Co: A Collaborative Test-time Refinement Framework for In-the-Wild 3D Human-Body Reconstruction Task](https://mlsanity.com/paper/2310.01291/) (2023)
- [3D Human Mesh Estimation from Single View RGBD](https://mlsanity.com/paper/2508.08178/) (2025)
- [Robust Long-term Test-Time Adaptation for 3D Human Pose Estimation through Motion Discretization](https://mlsanity.com/paper/2511.18851/) (2025)

## Abstract (arXiv)

We consider a new problem of adapting a human mesh reconstruction model to out-of-domain streaming videos, where performance of existing SMPL-based models are significantly affected by the distribution shift represented by different camera parameters, bone lengths, backgrounds, and occlusions. We tackle this problem through online adaptation, gradually correcting the model bias during testing. There are two main challenges: First, the lack of 3D annotations increases the training difficulty and results in 3D ambiguities. Second, non-stationary data distribution makes it difficult to strike a balance between fitting regular frames and hard samples with severe occlusions or dramatic changes. To this end, we propose the Dynamic Bilevel Online Adaptation algorithm (DynaBOA). It first introduces the temporal constraints to compensate for the unavailable 3D annotations, and leverages a bilevel optimization procedure to address the conflicts between multi-objectives. DynaBOA provides additional 3D guidance by co-training with similar source examples retrieved efficiently despite the distribution shift. Furthermore, it can adaptively adjust the number of optimization steps on individual frames to fully fit hard samples and avoid overfitting regular frames. DynaBOA achieves state-of-the-art results on three out-of-domain human mesh reconstruction benchmarks.