Research map / Video / Video segmentation and tracking
Optimizing Video Prediction via Video Frame Interpolation
Yue Wu, Qiang Wen, Qifeng Chen
Check 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 |
|---|---|---|---|
| Use VFI-constrained optimization without external video-prediction training | Video prediction · Cityscapes, t+1 | MS-SSIM (×1e-2): 89.10 -> 94.54 | Improved |
| Use VFI-constrained optimization without external video-prediction training | Video prediction · KITTI, t+1 | LPIPS (×1e-2): 18.48 -> 12.34 | Improved |
| Use VFI-constrained optimization without external video-prediction training | Video prediction · DAVIS, t+1 | LPIPS (×1e-2): 13.09 -> 11.40 | Improved |
| Use VFI-constrained optimization without external video-prediction training | Video prediction · Middlebury, t+1 | LPIPS (×1e-2): 7.98 -> 6.07 | Improved |
| Use VFI-constrained optimization without external video-prediction training | Video prediction · Vimeo90K, t+1 | LPIPS (×1e-2): 7.73 -> 3.59 | Improved |
| Add the image loss to the consistency loss (ablation) | Full video-prediction method · Cityscapes, t+1 | SSIM: 0.8939 -> 0.8975 | Improved |
| Add the consistency loss to the image loss (ablation) | Full video-prediction method · Cityscapes, t+1 | SSIM: 0.8732 -> 0.8975 | Improved |
| Replace L1 image loss with MSE (ablation) | Full video-prediction method · Cityscapes, t+1 | SSIM: 0.8975 -> 0.8877 | Hurt |
| Use the VFI intermediate output instead of the final interpolation result (ablation) | Full video-prediction method · Cityscapes, t+1 | SSIM: 0.8975 -> 0.8963 | No clear effect |
| Add a long-term image constraint using four input frames (ablation) | Full video-prediction method · Cityscapes, t+1 | SSIM: 0.8975 -> 0.6978 | Hurt |
| Add optical-flow inpainting (ablation) | Full video-prediction method · Cityscapes, t+1 | SSIM: 0.8882 -> 0.8975 | Improved |
| Initialize optical flow with zero values or Gaussian noise instead of a motion-based initialization (ablation) | Full video-prediction method · Cityscapes, t+1 | SSIM: 0.8975 -> 0.7719; 0.8975 -> 0.7669 | 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.
- DYAN — DYAN: A Dynamical Atoms-Based Network For Video Prediction (ECCV 2018) · 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
- Video Frame Interpolation without Temporal Priors (NeurIPS 2020 · similarity 0.88)
- Scene-Adaptive Video Frame Interpolation via Meta-Learning (CVPR 2020 · similarity 0.88)
- Super SloMo: High Quality Estimation of Multiple Intermediate Frames for Video Interpolation (CVPR 2018 · similarity 0.88)
- Video Frame Prediction from a Single Image and Events (AAAI 2024 · similarity 0.87)
- Repurposing Pre-trained Video Diffusion Models for Event-based Video Interpolation (CVPR 2025 · similarity 0.86)
- Generative Inbetweening: Adapting Image-to-Video Models for Keyframe Interpolation (ICLR 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.
