Research map / Video / Video segmentation and tracking
A Perceptual Quality Metric for Video Frame Interpolation
Qiqi Hou, Abhijay Ghildyal, Feng Liu
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 |
|---|---|---|---|
| use spatio-temporal video metric instead of per-frame image metric LPIPS | VFI perceptual metric · VFIPS validation | 2AFC: 0.808 -> 0.830 | Improved |
| use spatio-temporal video metric instead of per-frame LPIPS | VFI perceptual metric · BVI-VFI test | SROCC: 0.628 -> 0.794 | Improved |
| use our metric instead of strongest video baseline | VFI perceptual metric · VFIPS validation and BVI-VFI | 2AFC / SROCC / PLCC / KROCC: 0.025 in terms of 2AFC, 0.180 in SROCC, 0.188 in PLCC, 0.161 in KROCC | Improved |
| use our metric on large-motion 4K videos | VFI perceptual metric · X-TEST 4K | 2AFC: 0.748 -> 0.789 | Improved |
| use 2D per-frame extractor instead of AlexNet (ablation) | VFI perceptual metric · BVI-VFI test | SROCC: 0.761 -> 0.794 | Improved |
| use 2D extractor instead of I3D (ablation) | VFI perceptual metric · BVI-VFI test | SROCC: 0.659 -> 0.794 | Improved |
| use 2D extractor instead of Conv3D variant (Ours-3D) (ablation) | VFI perceptual metric · BVI-VFI test | SROCC: 0.728 -> 0.794 | Improved |
| add spatio-temporal module (Conv3D) over no module (ablation) | VFI perceptual metric · BVI-VFI test | SROCC: 0.617 -> 0.761 | Improved |
| use LN-free compact Swin ST module instead of Conv3D (ablation) | VFI perceptual metric · BVI-VFI test | SROCC: 0.761 -> 0.794 | Improved |
| remove LayerNorm layers from Swin block (ablation) | VFI perceptual metric · BVI-VFI test | SROCC / PLCC / KROCC: 0.070 on SROCC, 0.124 on PLCC, and 0.089 on KROCC | Improved |
| use original Swin block instead of Conv3D (ablation) | VFI perceptual metric · BVI-VFI test | SROCC: 0.761 -> 0.728 | Hurt |
| add automatic LPIPS annotations to human annotations (ablation) | VFI perceptual metric · BVI-VFI test | SROCC: 0.719 -> 0.794 | Improved |
Closest papers by meaning
- LAVIB: A Large-scale Video Interpolation Benchmark (NeurIPS 2024 · similarity 0.88)
- Perception-Oriented Video Frame Interpolation via Asymmetric Blending (CVPR 2024 · similarity 0.88)
- Super SloMo: High Quality Estimation of Multiple Intermediate Frames for Video Interpolation (CVPR 2018 · similarity 0.87)
- Depth-Aware Video Frame Interpolation (CVPR 2019 · similarity 0.86)
- KVQ: Boosting Video Quality Assessment via Saliency-guided Local Perception (CVPR 2025 · similarity 0.86)
- Disentangled Motion Modeling for Video Frame Interpolation (AAAI 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.
