Research map / Video / Video segmentation and tracking
Conditional Temporal Variational AutoEncoder for Action Video Prediction
Xiaogang Xu, Yi Wang, Liwei Wang, Bei Yu, Jiaya Jia
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 action labels to ACT-VAE (ablation) | ACT-VAE · Penn-action keypoints | L2: 29.59 -> 28.32 | Improved |
| add action labels to ACT-VAE (ablation) | ACT-VAE · Human3.6M keypoints | L2: 30.88 -> 30.41 | Improved |
| condition latent z_t on previous latents z_<t (ablation) | ACT-VAE · Penn-action keypoints | L2: 30.31 -> 29.59 | Improved |
| condition latent z_t on previous latents z_<t (ablation) | ACT-VAE · Penn-action keypoints | Std: 1.462 -> 1.584 | Improved |
| add temporal coherence over latents (ablation) | ACT-VAE · Penn-action keypoints | L2: 34.74 -> 30.31 | Improved |
| add temporal coherence over latents (ablation) | ACT-VAE · Penn-action keypoints | Std: 0.732 -> 1.462 | Improved |
| add foreground-background mask in P2I (ablation) | P2I network · Penn-action video | FVD: 1455.2 -> 1092.8 | Improved |
| add action conditional BN in P2I (ablation) | P2I network · Penn-action video | FVD: 1377.2 -> 1092.8 | Improved |
| add action conditional BN in P2I (ablation) | P2I network · Penn-action video | Recognition accuracy: 68.83 -> 70.04 | Improved |
| lower lambda_1 from 200 to 20 (ablation) | ACT-VAE · Penn-action keypoints | L2: 28.32 -> 31.84 | Hurt |
| raise lambda_2 from 0.002 to 0.02 (ablation) | ACT-VAE · Penn-action keypoints | L2: 28.32 -> 31.99 | Hurt |
| use ACT-VAE pose generator in place of VAE pose generator | P2I framework · Penn-action video | FVD: 1318.9 -> 1092.8 | Improved |
Closest papers by meaning
- Action-Conditioned 3D Human Motion Synthesis With Transformer VAE (ICCV 2021 · similarity 0.90)
- A Variational Auto-Encoder Model for Stochastic Point Processes (CVPR 2019 · similarity 0.90)
- Weakly-Supervised Action Localization by Generative Attention Modeling (CVPR 2020 · similarity 0.87)
- CV-VAE: A Compatible Video VAE for Latent Generative Video Models (NeurIPS 2024 · similarity 0.86)
- Improved Video VAE for Latent Video Diffusion Model (CVPR 2025 · similarity 0.86)
- Latent Action Pretraining from Videos (ICLR 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.
