MotionBERT: A Unified Perspective on Learning Human Motion Representations
We present a unified perspective on tackling various human-centric video tasks by learning human motion representations from large-scale and heterogeneous data resources. Specifically, we propose a pretraining stage in which a motion encoder is trained to recover the underlying 3D motion from noisy partial 2D observations. The motion representations acquired in this way incorporate geometric, kinematic, and physical knowledge about human motion, which can be easily transferred to multiple downstream tasks. We implement the motion encoder with a Dual-stream Spatio-temporal Transformer (DSTformer) neural network. It could capture long-range spatio-temporal relationships among the skeletal joints comprehensively and adaptively, exemplified by the lowest 3D pose estimation error so far when trained from scratch. Furthermore, our proposed framework achieves state-of-the-art performance on all three downstream tasks by simply finetuning the pretrained motion encoder with a simple regression head (1-2 layers), which demonstrates the versatility of the learned motion representations. Code and models are available at https://motionbert.github.io/
PDF Abstract ICCV 2023 PDF ICCV 2023 AbstractCode
MS COCO
Human3.6M
NTU RGB+D
3DPW
AMASS
NTU RGB+D 120
Results from the Paper
Ranked #1 on
Monocular 3D Human Pose Estimation
on Human3.6M
(using extra training data)
