TEMOS: Generating Diverse Human Motions from Text

Institute Homepage

Institute Homepage Sign In

Back

Research Overview

Inferring and exploiting contact

Generative Proxemics: A Prior for 3D Social Interaction from Images

BITE -- Dog Shape and Pose from an Image

HOLD -- inferring 3D hand and object shape from video

MOVER -- Reconstructing 3D Scenes and People using Interaction

Datasets for understanding humans and animals

The Poses for Equine Research Dataset (PFERD)

BEAT2 Dataset for Holistic Co-Speech Gesture Generation

ARCTIC: A Dataset for Dexterous Bimanual Hand-Object Manipulation

The BioAMASS Dataset

OpenCapBench dataset

Human health and the 3D body

Body Shape Models in Treating Anorexia Nervosa

Customized Bone Plants for Humerus Shaft Fractures

Reconstructing Signing Avatars From Video Using Linguistic Priors

The AI animator

HAAR: Text-Conditioned Generative Model of 3D Strand-based Human Hairstyles

Gaussian Garments

PuzzleAvatar: Assembling 3D Avatars from Personal Albums

FLARE: Fast Learning of Animatable and Relightable Mesh Avatars

Language, Vision, and World Models

AWOL: Analysis WithOut synthesis using Language

Re-Thinking Inverse Graphics with Large Language Models

TeCH: Text-guided Reconstruction of Clothed Humans

Human pose, shape, and motion capture

WHAM: Reconstructing World-grounded Humans with Accurate 3D Motion

3D Human Pose Estimation via Intuitive Physics

Accurate 3D Body Shape Regression using Metric and Semantic Attributes

BEV

Generating human motion

Generating Human Interaction Motions in Scenes with Text Control

TEMOS: Generating Diverse Human Motions from Text

EMAGE: Full-body Gestures from Audio

TEACH: Temporal Action Compositions for 3D Humans

Robot Perception Group

AirCap: 3D Motion Capture

AirCap: Perception-Based Control

AirCapRL: Aerial Motion Capture Using Deep RL

Data Team

Lab Tours and Public Outreach

Collecting Data - From the Idea to the Publication

Capture Technologies Setup

Completed Projects

Human Pose, Shape and Action

3D Pose from Images

2D Pose from Images

Beyond Motion Capture

Action and Behavior

Body Perception

Body Applications

Pose and Motion Priors

Clothing Models (2011-2015)

Reflectance Filtering

Learning on Manifolds

Markerless Animal Motion Capture

Multi-Camera Capture

2D Pose from Optical Flow

Body Perception

Neural Prosthetics and Decoding

Part-based Body Models

Intrinsic Depth

Lie Bodies

Layers, Time and Segmentation

Understanding Action Recognition (JHMDB)

Intrinsic Video

Intrinsic Images

Action Recognition with Tracking

Neural Control of Grasping

Flowing Puppets

Faces

Deformable Structures

Model-based Anthropometry

Modeling 3D Human Breathing

Optical flow in the LGN

FlowCap

Smooth Loops from Unconstrained Video

PCA Flow

Efficient and Scalable Inference

Motion Blur in Layers

Facade Segmentation

Smooth Metric Learning

Robust PCA

3D Recognition

Object Detection

Perceiving Systems Members Publications Website

TEMOS: Generating Diverse Human Motions from Text

Temossab — Using a novel cross-modal variational model, TEMOS generates diverse human motions given a text description [].

We address the problem of generating diverse 3D human motions from textual descriptions. This challenging task requires joint modeling of both modalities: understanding and extracting useful human-centric information from the text, and then generating plausible and realistic sequences of human poses. In contrast to most previous work which focuses on generating a single, deterministic, motion from a textual description, we design a variational approach that can produce multiple diverse human motions. We propose TEMOS, a text-conditioned generative model leveraging variational autoencoder (VAE) training with human motion data, in combination with a text encoder that produces distribution parameters compatible with the VAE latent space. We show the TEMOS framework can produce both skeleton-based animations as in prior work, as well more expressive SMPL body motions. We evaluate our approach on the KIT Motion-Language benchmark and, despite being relatively straightforward, demonstrate significant improvements over the state of the art. Code and models are available on our webpage.