Research map / Video / Video segmentation and tracking
Extracting Motion and Appearance via Inter-Frame Attention for Efficient Video Frame Interpolation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 single-frame attention with inter-frame attention for appearance features (ablation) | EMA-VFI · Vimeo90K | PSNR: 35.54 -> 36.02 | Helped |
| replace single-frame attention plus bilateral cost volume with IFA for both appearance and motion (ablation) | EMA-VFI · Xiph-4K | PSNR: 33.34 -> 34.25 | Helped |
| use IFA instead of mixed attention across the two frames (ablation) | EMA-VFI · Xiph-4K | PSNR: 33.88 -> 34.25 | Helped |
| use motion features M_t rather than directly inputting timestep t as the cue (ablation) | EMA-VFI arbitrary-timestep interpolation · HD 720p | PSNR: 32.05 -> 32.17 | Helped |
| increase Transformer block counts from 2/2 to 4/4 while keeping C=16 (ablation) | EMA-VFI · Xiph-2K | PSNR: 36.55 -> 36.61 | Helped |
| increase Transformer channels from C=16 to C=32 while keeping block counts 2/2 (ablation) | EMA-VFI · Xiph-2K | PSNR: 36.55 -> 36.70 | Helped |
| use Transformers only at the lowest feature scale (ablation) | EMA-VFI hybrid architecture · Vimeo90K | PSNR: 36.07 -> 35.26 | Hurt |
| extend Transformer blocks to one additional higher feature scale (ablation) | EMA-VFI hybrid architecture · Vimeo90K | PSNR: 36.07 -> 36.10 | No clear effect |
| increase IFA window size from 7 to 9 (ablation) | EMA-VFI · Xiph-4K | PSNR: 34.25 -> 34.18 | Hurt |
| increase model size from Ours-small to Ours | EMA-VFI · SNU-FILM Easy | PSNR: 39.81 -> 39.98 | Hurt |
| use EMA-VFI rather than VFIFormer | video frame interpolation · Xiph-4K | PSNR: 33.99 -> 34.67 | Helped |
Closest papers by meaning
- Disentangled Motion Modeling for Video Frame Interpolation (AAAI 2025 · similarity 0.88)
- Perception-Oriented Video Frame Interpolation via Asymmetric Blending (CVPR 2024 · similarity 0.87)
- Sparse Global Matching for Video Frame Interpolation with Large Motion (CVPR 2024 · similarity 0.87)
- Repurposing Pre-trained Video Diffusion Models for Event-based Video Interpolation (CVPR 2025 · similarity 0.86)
- EDEN: Enhanced Diffusion for High-quality Large-motion Video Frame Interpolation (CVPR 2025 · similarity 0.86)
- IQ-VFI: Implicit Quadratic Motion Estimation for Video Frame Interpolation (CVPR 2024 · 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.