ChatPose: Chatting about 3D Human Pose

Yao Feng1,2,3, Jing Lin4, Sai Kumar Dwivedi1, Yu Sun3, Priyanka Patel 1, Michael J. Black 1
1Max Planck Institute for Intelligent Systems, 2ETH Zurich, 3Meshcapade, 4Tsinghua University
arXiv Code Dataset

CVPR 2024

We introduce ChatPose, a multi-model LLM designed for chatting about human pose that produces 3D human poses (SMPL pose parameters) upon user request. ChatPose features a specialized SMPL projection layer trained to convert language embeddings into 3D human pose parameters. Our demonstration includes conversations both without (left) and with (right) an image input. Upon detection of a pose token, the token is used to estimate the SMPL pose parameters and subsequently generate the corresponding 3D body mesh.

Abstract

WWe introduce ChatPose, a framework employing Large Language Models (LLMs) to understand and reason about 3D human poses from images or textual descriptions. Our work is motivated by the human ability to intuitively understand postures from a single image or a brief description, a process that intertwines image interpretation, world knowledge, and an understanding of body language. Traditional human pose estimation and generation methods often operate in isolation, lacking semantic understanding and reasoning abilities. ChatPose addresses these limitations by embedding SMPL poses as distinct signal tokens within a multimodal LLM, enabling the direct generation of 3D body poses from both textual and visual inputs. Leveraging the powerful capabilities of multimodal LLMs, ChatPose unifies classical 3D human pose and generation tasks while offering user interactions. Additionally, ChatPose empowers LLMs to apply their extensive world knowledge in reasoning about human poses, leading to two advanced tasks: speculative pose generation and reasoning about pose estimation. These tasks involve reasoning about humans to generate 3D poses from subtle text queries, possibly accompanied by images. We establish benchmarks for these tasks, moving beyond traditional 3D pose generation and estimation methods. Our results show that ChatPose outperforms existing multimodal LLMs and task-specific methods on these newly proposed tasks. Furthermore, ChatPose's ability to understand and generate 3D human poses based on complex reasoning opens new directions in human pose analysis.

Method Overview

Our model is composed of a multi-modal LLM (with vision encoder, vision projection layer and LLM), a SMPL projection layer, and the parametric human body model, i.e. SMPL. The multi-modal LLM processes text and image inputs (if provided) to generate textual responses. In the training phase, we focus on training the SMPL projection layer and fine-tuning the LLM, while keeping the other components frozen. The three data types used for the end-to-end training are: text-to-3D pose generation, image-to-pose estimation, and multi-modal instruction-following data. When an image is available, its information is used by the LLM to deduce an answer. If the user inquires about a SMPL pose, the LLM responds with a token. The embedding related to this token is then used to predict the SMPL pose parameters, leading to the generation of a body mesh, as visualized.

Results

Examples for Speculative Pose Genereation (SPG). The query examples are sourced from our SPG benchmark, which offers implicit text queries regarding human poses. GPT-4 (with DALL·E 2) generates images that depict the correct pose but does not explictly generate 3D poses. In contrast, PoseScript is a task-specific method for 3D pose from language but it is not able to relate high-level concepts like ``searching under furniture" with 3D pose. Our method ChatPose, understands high-level concepts and how to relate them to 3D pose.

Examples for Reasoning-based Pose Estimation (RPE). The queries are from our RPE benchmark, which offers different types of text descriptions of the target person. Comparison with LLaVA and classical HMR-style methods HMR2.0 and SPIN on RPE. LLaVA-S refers to the process of using LLaVA for keypoint detection, followed by SMPL pose optimization via SMPLify. LLaVA-P uses LLaVA to obtain textual descriptions of the pose, which are then input into PoseScript to generate human poses. For each method, we utilize the entire image provided by the user as input, without applying cropping. Methods involving LLMs are highlighted in orange, while those that are purely task-specific methods, are marked in green.

We provide the benchmark annotations here.

Acknowledgements

We thank Weiyang Liu, Haiwen Feng and Longhui Yu for discussions and proofreading. We also thank Naureen Mahmood and Nicolas Keller for support with data. This work was partially supported by the Max Planck ETH Center for Learning Systems.

Disclosure. MJB has received research gift funds from Adobe, Intel, Nvidia, Meta/Facebook, and Amazon. MJB has financial interests in Amazon and Meshcapade GmbH. While MJB is a co-founder and Chief Scientist at Meshcapade, his research in this project was performed solely at, and funded solely by, the Max Planck Society.


Related Links

Note: The original name of ChatPose, 'PoseGPT,' was changed to avoid confusion with similarly named previous works

For more related works, please check out the following links:

Third wave 3D human pose and shape estimation A blog about the development of 3D human pose and shape estimation.

PoseScript for 3D human pose generation from text.

HMR,SPIN,HMR2 for 3D human pose and shape estimation from images.

LLaVA,LISA,Next-GPT for multimodal LLMs.

BibTeX

    
@InProceedings{feng2024chatpose,
    author = {Feng, Yao and Lin, Jing and Dwivedi, Sai Kumar and Sun, Yu and Patel, Priyanka and Black, Michael J.},
    title = {ChatPose: Chatting about 3D Human Pose},
    booktitle = {CVPR},
    year = {2024}
}