Research map / Video / Video generation
Hybrid Video Diffusion Models with 2D Triplane and 3D Wavelet Representation
Kihong Kim, Haneol Lee, Jihye Park, Seyeon Kim, Kwang Hee Lee, Seungryong Kim, Jaejun Yoo
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 frequency matching loss in wavelet domain to autoencoder training (ablation) | HVDM autoencoder · UCF-101 reconstruction | R-FVD: 8.08 -> 5.35 | Improved |
| add frequency matching loss (ablation) | HVDM autoencoder · UCF-101 reconstruction | SSIM: 0.892 -> 0.915 | Improved |
| combine 2D triplane with 3D wavelet volume latent instead of triplane alone (no adversarial loss) (ablation) | HVDM autoencoder · UCF-101 reconstruction | R-FVD: 180.21 -> 5.35 | Improved |
| combine 2D triplane with 3D wavelet latent instead of 3D CNN alone (no adversarial loss) (ablation) | HVDM autoencoder · UCF-101 reconstruction | R-FVD: 86.73 -> 5.35 | Improved |
| add adversarial 3D discriminator to triplane autoencoder (ablation) | Video Transformer triplane autoencoder · UCF-101 reconstruction | R-FVD: 180.21 -> 27.03 (PSNR 28.12 -> 27.03) | Mixed |
| replace PVDM with HVDM for generation on UCF-101 long clips | HVDM · UCF-101 generation | FVD128: 505.0 -> 549.7 | Hurt |
| use HVDM hybrid latent instead of PVDM triplane latent | latent video diffusion · UCF-101 generation | FVD16: 399.4 -> 303.1 | Improved |
| use HVDM hybrid latent instead of PVDM triplane latent | latent video diffusion · TaiChi generation | FVD16: 267.0 -> 77.0 | Improved |
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.
- StyleGAN-V — StyleGAN-V: A Continuous Video Generator With the Price, Image Quality and Perks of StyleGAN2 (CVPR 2022) · better on 1 of 1 shared result
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
- Efficient Video Diffusion Models via Content-Frame Motion-Latent Decomposition (ICLR 2024 · similarity 0.90)
- World-consistent Video Diffusion with Explicit 3D Modeling (CVPR 2025 · similarity 0.88)
- OSV: One Step is Enough for High-Quality Image to Video Generation (CVPR 2025 · similarity 0.88)
- From Slow Bidirectional to Fast Autoregressive Video Diffusion Models (CVPR 2025 · similarity 0.87)
- VideoCrafter2: Overcoming Data Limitations for High-Quality Video Diffusion Models (CVPR 2024 · similarity 0.87)
- Hierarchical Patch Diffusion Models for High-Resolution Video Generation (CVPR 2024 · similarity 0.87)
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.
