AMASS: Archive of Motion Capture As Surface Shapes

By Naureen Mahmood, Nima Ghorbani, N. Troje, Gerard Pons-Moll, Michael J. Black. IEEE International Conference on Computer Vision, 2019

1,942 citations recorded by the source corpus.

Large datasets are the cornerstone of recent advances in computer vision using deep learning. In contrast, existing human motion capture (mocap) datasets are small and the motions limited, hampering progress on learning models of human motion. While there are many different datasets available, they each use a different parameterization of the body, making it difficult to integrate them into a single meta dataset. To address this, we introduce AMASS, a large and varied database of human motion that unifies 15 different optical marker-based mocap datasets by representing them within a common framework and parameterization. We achieve this using a new method, MoSh++, that converts mocap data into realistic 3D human meshes represented by a rigged body model. Here we use SMPL [Loper et al., 2015], which is widely used and provides a standard skeletal representation as well as a fully rigged surface mesh. The method works for arbitrary marker sets, while recovering soft-tissue dynamics and realistic hand motion. We evaluate MoSh++ and tune its hyperparameters using a new dataset of 4D body scans that are jointly recorded with markerbased mocap. The consistent representation of AMASS makes it readily useful for animation, visualization, and generating training data for deep learning. Our dataset is significantly richer than previous human motion collections, having more than 40 hours of motion data, spanning over 300 subjects, more than 11000 motions, and will be publicly available to the research community.

Subfields: Computational Mechanics, Computer Vision and Pattern Recognition, Control and Systems Engineering

View Paper: https://www.semanticscholar.org/paper/690c817ab5be8017ff713fa1028669debde205af

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Mechanics

In Computer Vision and Pattern Recognition

In Control and Systems Engineering

Citations in the corpus

MoMask: Generative Masked Modeling of 3D Human Motions (2023, methodology): “HumanML3D dataset collects 14,616 motions from AMASS [32] and HumanAct12 [14] datasets, with each motion described by 3 textual scripts, totaling 44,970 descriptions.”
Pose2Mesh: Graph Convolutional Network for 3D Human Pose and Mesh Recovery from a 2D Human Pose (2020, methodology): “We leverage additional synthetic data from AMASS [37] to boost the performance of Pose2Mesh.”
CLIFF: Carrying Location Information in Full Frames into Human Pose and Shape Estimation (2022, neutral): “Compared with other annotators whose priors come from extra models [4,41] trained on another large motion capture dataset AMASS [31], the CLIFF annotator contains strong priors that are efficient to obtain with no need for an extra model and AMASS.”
PointOdyssey: A Large-Scale Synthetic Dataset for Long-Term Point Tracking (2023, methodology): “To animate the humanoid characters, we use real-world human motion data [38, 67, 66].”
Human Motion Diffusion Model (2022, methodology): “HumanML3D is a recent dataset, textually re-annotating motion capture from the AMASS (Mahmood et al., 2019) and HumanAct12 (Guo et al.”
MotionCLIP: Exposing Human Motion Generation to CLIP Space (2022, methodology): “More recently, BABEL dataset [37] provided per-frame textual labels ordered in 260 classes to the larger AMASS dataset [29], including about 40 hours of motion capture.”
On Self-Contact and Human Pose (2021, methodology): “Sampling meshes from AMASS.”
AI Choreographer: Music Conditioned 3D Dance Generation with AIST++ (2021, background): “While there are many large scale 3D motion capture datasets [39, 59, 1, 37], mocap dataset of 3D dance is quite limited as it requires heavy instrumentation and expert dancers for capture.”
Deep 3D human pose estimation: A review (2021, neutral): “For example, Choutas et al. (2020) propose (2019)) together with unpaired, in-the-wild, 2D keypoint annotations, a video inference for body pose and shape estimation (VIBE) is proposed in Kocabas et al. (2020).”
History Repeats Itself: Human Motion Prediction via Motion Attention (2020, methodology): “The results of short-term and long-term prediction in 3D on AMASS and 3DPW are shown in Table 5.”
MotionGPT: Human Motion as a Foreign Language (2023, positive): “The KIT dataset provides 6,353 textual descriptions corresponding to 3,911 motion sequences, while the HumanML3D dataset [10] is a more recent dataset that contains 14,616 motion sequences obtained from AMASS [25], along with 44,970 sequence-level textual descriptions.”
BABEL: Bodies, Action and Behavior with English Labels (2021, positive): “We collect BABEL by showing rendered videos of mo-cap sequences from AMASS [22] to human annotators and eliciting action labels (Fig.”
Perpetual Humanoid Control for Real-time Simulated Avatars (2023, methodology): “PHC is trained on the training split of the AMASS [23] dataset.”
Generating Diverse and Natural 3D Human Motions from Text (2022, methodology): “Our HumanML3D dataset originates from a amalgamation of motion sequences from the HumanAct12 [12] and AMASS [23] datasets, two large-scale datasets of 3D human motion captures that are publicly accessible.”
MotionDiffuse: Text-Driven Human Motion Generation With Diffusion Model (2022, methodology): “HumanML3D [48] re-annotates the AMASS dataset [73] and the HumanAct12 dataset [1].”
HumanPlus: Humanoid Shadowing and Imitation from Humans (2024, methodology): “At the same time, we have a 40-hour human motion dataset, AMASS [49], that covers a wide range of skills.”
Deep Learning-based Human Pose Estimation: A Survey (2020, methodology): “X, which is an improved version learned from AMASS dataset [174].”