Controllable Motion Synthesis and Reconstruction with Autoregressive Diffusion Models

被引:0
|
作者
Yin, Wenjie [1 ]
Tu, Ruibo [1 ]
Yin, Hang [1 ]
Kragic, Danica [1 ]
Kjellstrom, Hedvig [1 ]
Bjorkman, Marten [1 ]
机构
[1] KTH Royal Inst Technol, Div Robot Percept & Learning, Stockholm, Sweden
基金
欧盟地平线“2020”; 瑞典研究理事会;
关键词
D O I
10.1109/RO-MAN57019.2023.10309317
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Data-driven and controllable human motion synthesis and prediction are active research areas with various applications in interactive media and social robotics. Challenges remain in these fields for generating diverse motions given past observations and dealing with imperfect poses. This paper introduces MoDiff, an autoregressive probabilistic diffusion model over motion sequences conditioned on control contexts of other modalities. Our model integrates a cross-modal Transformer encoder and a Transformer-based decoder, which are found effective in capturing temporal correlations in motion and control modalities. We also introduce a new data dropout method based on the diffusion forward process to provide richer data representations and robust generation. We demonstrate the superior performance of MoDiff in controllable motion synthesis for locomotion with respect to two baselines and show the benefits of diffusion data dropout for robust synthesis and reconstruction of high-fidelity motion close to recorded data.
引用
收藏
页码:1102 / 1108
页数:7
相关论文
共 50 条
  • [1] Guided Motion Diffusion for Controllable Human Motion Synthesis
    Karunratanakul, Korrawe
    Preechakul, Konpat
    Suwajanakorn, Supasorn
    Tang, Siyu
    [J]. 2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION, ICCV, 2023, : 2151 - 2162
  • [2] A Survey of Multimodal Controllable Diffusion Models
    Jiang, Rui
    Zheng, Guang-Cong
    Li, Teng
    Yang, Tian-Rui
    Wang, Jing-Dong
    Li, Xi
    [J]. JOURNAL OF COMPUTER SCIENCE AND TECHNOLOGY, 2024, 39 (03) : 509 - 541
  • [3] Autoregressive Stylized Motion Synthesis with Generative Flow
    Wen, Yu-Hui
    Yang, Zhipeng
    Fu, Hongbo
    Gao, Lin
    Sun, Yanan
    Liu, Yong-Jin
    [J]. 2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021, 2021, : 13607 - 13616
  • [4] Controllable motion synthesis in a gaseous medium
    Shi, L
    Yu, YZ
    Wojtan, C
    Chenney, S
    [J]. VISUAL COMPUTER, 2005, 21 (07): : 474 - 487
  • [5] Controllable motion synthesis in a gaseous medium
    Lin Shi
    Yizhou Yu
    Christopher Wojtan
    Stephen Chenney
    [J]. The Visual Computer, 2005, 21 : 474 - 487
  • [6] Tracking articulated motion using a mixture of autoregressive models
    Agarwal, A
    Triggs, B
    [J]. COMPUTER VISION - ECCV 2004, PT 3, 2004, 3023 : 54 - 65
  • [7] Controllable Variation Synthesis for Surface Motion Capture
    Boukhayma, Adnane
    Boyer, Edmond
    [J]. PROCEEDINGS 2017 INTERNATIONAL CONFERENCE ON 3D VISION (3DV), 2017, : 309 - 317
  • [8] Autoregressive Models for Statistical Parametric Speech Synthesis
    Shannon, Matt
    Zen, Heiga
    Byrne, William
    [J]. IEEE TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING, 2013, 21 (03): : 587 - 597
  • [9] Optimized order estimation for autoregressive models to predict respiratory motion
    Robert Dürichen
    Tobias Wissel
    Achim Schweikard
    [J]. International Journal of Computer Assisted Radiology and Surgery, 2013, 8 : 1037 - 1042
  • [10] Optimized order estimation for autoregressive models to predict respiratory motion
    Duerichen, Robert
    Wissel, Tobias
    Schweikard, Achim
    [J]. INTERNATIONAL JOURNAL OF COMPUTER ASSISTED RADIOLOGY AND SURGERY, 2013, 8 (06) : 1037 - 1042