Research map / Video / Video generation
Disentangled Hierarchical VAE for 3D Human-Human Interaction Generation
Zichen Geng, Zeeshan Hayder, Bo Miao, Jian Liu, Wei Liu, Ajmal Saeed Mian
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add triplet-loss contrastive learning on z_o (ablation) | DHVAE + denoiser · InterHuman | gFID: 5.063 -> 5.015 | Improved |
| add triplet-loss contrastive learning on z_o (ablation) | DHVAE + denoiser · InterHuman | rFID: 0.499 -> 0.503 | No clear effect |
| add triplet-loss contrastive learning on z_o (ablation) | DHVAE · InterHuman penetration | Penetration Volume: 0.446 -> 0.390 | Improved |
| add global interaction latent z_o (ablation) | DHVAE · InterHuman | gFID: 5.886 -> 5.015 | Improved |
| add CoTransformer (ablation) | DHVAE · InterHuman | gFID: 5.593 -> 5.015 | Improved |
| use disentangled hierarchical VAE instead of MLD-VAE (ablation) | DHVAE · InterHuman | gFID: 5.694 -> 5.015 | Improved |
| use DHVAE instead of MLD-VAE for reconstruction | DHVAE · InterHuman reconstruction | rFID: 1.011 -> 0.503 | Improved |
| add structural positional embedding (SPE) in denoiser (ablation) | Denoiser · InterHuman | gFID: 7.452 -> 5.015 | Improved |
| add token scaling in denoiser (ablation) | Denoiser · InterHuman | gFID: 6.531 -> 5.015 | Improved |
| add skip connections in denoiser (ablation) | Denoiser · InterHuman | gFID: 5.343 -> 5.015 | Improved |
| increase individual encoder layers from 4 to 8 (3 CoTransformer layers) (ablation) | DHVAE · InterHuman | FID: 5.015 -> 5.612 | Hurt |
| increase latent size l from 1 to 7 (ablation) | DHVAE · InterHuman | FID: 5.015 -> 5.670 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- TIMotion* — TIMotion: Temporal and Interactive Framework for Efficient Human-Human Motion Generation (CVPR 2025) · better on 6 of 6 shared results
- InterMask — InterMask: 3D Human Interaction Generation via Collaborative Masked Modeling (ICLR 2025) · better on 6 of 6 shared results
- InterGen — InterGen: Diffusion-Based Multi-human Motion Generation Under Complex Interactions (IJCV 2024) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Auto-Regressive Diffusion for Generating 3D Human-Object Interactions (AAAI 2025 · similarity 0.88)
- CG-HOI: Contact-Guided 3D Human-Object Interaction Generation (CVPR 2024 · similarity 0.87)
- InterDreamer: Zero-Shot Text to 3D Dynamic Human-Object Interaction (NeurIPS 2024 · similarity 0.87)
- InterMask: 3D Human Interaction Generation via Collaborative Masked Modeling (ICLR 2025 · similarity 0.86)
- HOI-Dyn: Learning Interaction Dynamics for Human-Object Motion Diffusion (NeurIPS 2025 · similarity 0.86)
- InterAct: Advancing Large-Scale Versatile 3D Human-Object Interaction Generation (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
