Papers › Action-conditioned On-demand Motion Generation

Action-conditioned On-demand Motion Generation

17 Jul 2022arXiv:2207.08164archive 2025-07-28

QIUJING LU, YiPeng Zhang, Mingjian Lu, Vwani Roychowdhury

We propose a novel framework, On-Demand MOtion Generation (ODMO), for generating realistic and diverse long-term 3D human motion sequences conditioned only on action types with an additional capability of customization. ODMO shows improvements over SOTA approaches on all traditional motion evaluation metrics when evaluated on three public datasets (HumanAct12, UESTC, and MoCap). Furthermore, we provide both qualitative evaluations and quantitative metrics demonstrating several first-known customization capabilities afforded by our framework, including mode discovery, interpolation, and trajectory customization. These capabilities significantly widen the spectrum of potential applications of such motion generation models. The novel on-demand generative capabilities are enabled by innovations in both the encoder and decoder architectures: (i) Encoder: Utilizing contrastive learning in low-dimensional latent space to create a hierarchical embedding of motion sequences, where not only the codes of different action types form different groups, but within an action type, codes of similar inherent patterns (motion styles) cluster together, making them readily discoverable; (ii) Decoder: Using a hierarchical decoding strategy where the motion trajectory is reconstructed first and then used to reconstruct the whole motion sequence. Such an architecture enables effective trajectory control. Our code is released on the Github page: https://github.com/roychowdhuryresearch/ODMO

PaperPDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

roychowdhuryresearch/odmo officialmentioned in paperpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Contrastive LearningDecoderHuman action generationMotion Generation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Human action generation CMU Mocap ODMO Accuracy 93.51 #1 of 1 Archive leaderboard report
Human action generation CMU Mocap ODMO Diversity 6.56 #1 of 1 Archive leaderboard report
Human action generation CMU Mocap ODMO FID 34 #1 of 1 Archive leaderboard report
Human action generation CMU Mocap ODMO Multimodality 2.49 #1 of 1 Archive leaderboard report
Human action generation HumanAct12 ODMO Accuracy 97.81 #1 of 1 Archive leaderboard report
Human action generation HumanAct12 ODMO Diversity 0.705 #1 of 1 Archive leaderboard report
Human action generation HumanAct12 ODMO FID 0.12 #1 of 1 Archive leaderboard report
Human action generation HumanAct12 ODMO Multimodality 2.57 #1 of 1 Archive leaderboard report
Human action generation UESTC RGB-D ODMO Accuracy 93.67 #1 of 1 Archive leaderboard report
Human action generation UESTC RGB-D ODMO Diversity 7.11 #1 of 1 Archive leaderboard report
Human action generation UESTC RGB-D ODMO FID 0.15 #1 of 1 Archive leaderboard report
Human action generation UESTC RGB-D ODMO Test 0.17 #1 of 1 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Contrastive Learning

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections