Research map / Video / Video generation
HERO: Human Reaction Generation from Videos
Chengjun Yu, Wei Zhai, Yuhang Yang, Yang Cao, Zheng-Jun Zha
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 Interaction Intention Extraction (global-local cross-attention) (ablation) | HERO · ViMo test set | FID: 0.535 -> 0.427 | Improved |
| add Interaction Intention Extraction (ablation) | HERO · ViMo test set | MModality: 1.626 -> 1.614 | No clear effect |
| add Dynamic Information Exploitation (motion-frame cross-attention) (ablation) | HERO · ViMo test set | FID: 0.521 -> 0.427 | Improved |
| add Dynamic Information Exploitation (ablation) | HERO · ViMo test set | MModality: 1.443 -> 1.614 | Improved |
| use 6 Transformer decoder units instead of 4 (ablation) | HERO · ViMo test set | FID: 0.659 -> 0.427 | Improved |
| use 6 Transformer decoder units instead of 8 (ablation) | HERO · ViMo test set | FID: 0.683 -> 0.427 | Improved |
| use 6 Transformer decoder units instead of 7 (ablation) | HERO · ViMo test set | MModality: 1.614 -> 1.649 | No clear effect |
| increase training pairs from 2800 to 8400 (ablation) | HERO · ViMo test set | FID: 0.427 -> 0.376 | Improved |
| increase training pairs from 2800 to 8400 (ablation) | HERO · ViMo test set | MModality: 1.614 -> 1.422 | Hurt |
| replace MoMask with HERO on unseen categories | HERO vs MoMask · ViMo Unseen set | FID: 2.477 -> 2.156 | Improved |
| replace MoMask with HERO | HERO vs MoMask · ViMo test set | FID: 0.856 -> 0.427 | Improved |
| replace MoMask with HERO | HERO vs MoMask · ViMo test set | MModality: 1.567 -> 1.614 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BAMM — BAMM: Bidirectional Autoregressive Motion Model (ECCV 2024) · better on 2 of 3 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
- ReGenNet: Towards Human Action-Reaction Synthesis (CVPR 2024 · similarity 0.88)
- InterDreamer: Zero-Shot Text to 3D Dynamic Human-Object Interaction (NeurIPS 2024 · similarity 0.87)
- HumanVid: Demystifying Training Data for Camera-controllable Human Image Animation (NeurIPS 2024 · similarity 0.86)
- InterAct: Advancing Large-Scale Versatile 3D Human-Object Interaction Generation (CVPR 2025 · similarity 0.85)
- Ready-to-React: Online Reaction Policy for Two-Character Interaction Generation (ICLR 2025 · similarity 0.85)
- Learning to Generate Human-Human-Object Interactions from Textual Descriptions (NeurIPS 2025 · similarity 0.85)
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.
