# Amortized Inverse Kinematics via Graph Attention for Real-Time Human Avatar Animation

arXiv: https://arxiv.org/abs/2604.16629 · HTML: https://mlsanity.com/paper/2604.16629/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 20 of its 43 references that are in the map.
- First posted on arXiv 17 Apr 2026.

## Builds on (most cited first, 20 of 20)

- [Graph Attention Networks](https://mlsanity.com/paper/1710.10903/) (2017)
- [Expressive Body Capture: 3D Hands, Face, and Body from a Single Image](https://mlsanity.com/paper/1904.05866/) (2019)
- [AMASS: Archive of Motion Capture as Surface Shapes](https://mlsanity.com/paper/1904.03278/) (2019)
- [On the Continuity of Rotation Representations in Neural Networks](https://mlsanity.com/paper/1812.07035/) (2018)
- [Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition](https://mlsanity.com/paper/1801.07455/) (2018)
- [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)
- [Humans in 4D: Reconstructing and Tracking Humans with Transformers](https://mlsanity.com/paper/2305.20091/) (2023)
- [HybrIK: A Hybrid Analytical-Neural Inverse Kinematics Solution for 3D Human Pose and Shape Estimation](https://mlsanity.com/paper/2011.14672/) (2020)
- [Semantic Graph Convolutional Networks for 3D Human Pose Regression](https://mlsanity.com/paper/1904.03345/) (2019)
- [Multi-HMR: Multi-Person Whole-Body Human Mesh Recovery in a Single Shot](https://mlsanity.com/paper/2402.14654/) (2024)
- [UnrealEgo: A New Dataset for Robust Egocentric 3D Human Motion Capture](https://mlsanity.com/paper/2208.01633/) (2022)
- [Deep Inertial Poser: Learning to Reconstruct Human Pose from Sparse Inertial Measurements in Real Time](https://mlsanity.com/paper/1810.04703/) (2018)
- [NIKI: Neural Inverse Kinematics with Invertible Neural Networks for 3D Human Pose and Shape Estimation](https://mlsanity.com/paper/2305.08590/) (2023)
- [DiffusionPoser: Real-time Human Motion Reconstruction From Arbitrary Sparse Sensors Using Autoregressive Diffusion](https://mlsanity.com/paper/2308.16682/) (2023)
- [SparsePoser: Real-time Full-body Motion Reconstruction from Sparse Data](https://mlsanity.com/paper/2311.02191/) (2023)
- [Neural Inverse Kinematics](https://mlsanity.com/paper/2205.10837/) (2022)
- [Quater-GCN: Enhancing 3D Human Pose Estimation with Orientation and Semi-supervised Training](https://mlsanity.com/paper/2404.19279/) (2024)
- [KinePose: A temporally optimized inverse kinematics technique for 6DOF human pose estimation with biomechanical constraints](https://mlsanity.com/paper/2207.12841/) (2022)
- [SIMSPINE: A Biomechanics-Aware Simulation Framework for 3D Spine Motion Annotation and Benchmarking](https://mlsanity.com/paper/2602.20792/) (2026)

## Abstract (arXiv)

Inverse kinematics (IK) is a core operation in animation, robotics, and biomechanics: given Cartesian constraints, recover joint rotations under a known kinematic tree. In many real-time human avatar pipelines, the available signal per frame is a sparse set of tracked 3D joint positions, whereas animation systems require joint orientations to drive skinning. Recovering full orientations from positions is underconstrained, most notably because twist about bone axes is ambiguous, and classical IK solvers typically rely on iterative optimization that can be slow and sensitive to noisy inputs. We introduce IK-GAT, a lightweight graph-attention network that reconstructs full-body joint orientations from 3D joint positions in a single forward pass. The model performs message passing over the skeletal parent-child graph to exploit kinematic structure during rotation inference. To simplify learning, IK-GAT predicts rotations in a bone-aligned world-frame representation anchored to rest-pose bone frames. This parameterization makes the twist axis explicit and is exactly invertible to standard parent-relative local rotations given the kinematic tree and rest pose. The network uses a continuous 6D rotation representation and is trained with a geodesic loss on SO(3) together with an optional forward-kinematics consistency regularizer. IK-GAT produces animation-ready local rotations that can directly drive a rigged avatar or be converted to pose parameters of SMPL-like body models for real-time and online applications. With 374K parameters and over 650 FPS on CPU, IK-GAT outperforms VPoser-based per-frame iterative optimization without warm-start at significantly lower cost, and is robust to initial pose and input noise