Research map / Video / Video generation
Diverse Generation from a Single Video Made Possible
Niv Haim, Ben Feinstein, Niv Granot, Assaf Shocher, Shai Bagon, Tali Dekel, Michal Irani
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 |
|---|---|---|---|
| replace HP-VAE-GAN with patch nearest-neighbor generation (VGPNN) | single-video generation · HP-VAE-GAN dataset (10 videos) | SVFID: 0.0081 -> 0.0072 | Improved |
| replace SinGAN-GIF with patch nearest-neighbor generation (VGPNN) | single-video generation · SinGAN-GIF dataset (5 videos) | SVFID: 0.0119 -> 0.0058 | Improved |
| replace GAN training with non-parametric patch search | single-video generation · HP-VAE-GAN dataset (10 videos) | runtime: 7.625 days -> 18 secs | Improved |
| VGPNN versus HP-VAE-GAN in a user study | VGPNN · HP-VAE-GAN dataset (10 videos) | user preference (%): 67.84 | Improved |
| VGPNN versus SinGAN-GIF in a user study | VGPNN · SinGAN-GIF dataset (5 videos) | user preference (%): 50.57 | No clear effect |
| use 1 EM-like iteration and patch size (3x5x5) for videos above 3,000,000 voxels (ablation) | VGPNN · large videos | quality/runtime: 5 -> 1 iterations | No clear effect |
| use nearest instead of cubic interpolation for temporal downscaling (ablation) | VGPNN · pyramid construction | visual quality | No clear effect |
| keep dynamic-structure channels in Q and K at all scales (ablation) | video analogies · video analogies | layout vs appearance | Mixed |
Closest papers by meaning
- Drop the GAN: In Defense of Patches Nearest Neighbors As Single Image Generative Models (CVPR 2022 · similarity 0.91)
- Hierarchical Patch VAE-GAN: Generating Diverse Videos from a Single Sample (NeurIPS 2020 · similarity 0.90)
- Towards Smooth Video Composition (ICLR 2023 · similarity 0.90)
- Temporal Generative Adversarial Nets With Singular Value Clipping (ICCV 2017 · similarity 0.88)
- SnapGen-V: Generating a Five-Second Video within Five Seconds on a Mobile Device (CVPR 2025 · similarity 0.87)
- SF-V: Single Forward Video Generation Model (NeurIPS 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.
