Research map / Video / Video segmentation and tracking
MASS: Motion-Aligned Selective Scan for Flow-Based Video Frame Interpolation
Jun-Sang YOO, Seung-Won Jung
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 learnable non-linear path (residual velocity) to linear trajectory (ablation) | MASS-S · SNU-FILM Extreme | PSNR: 25.49 -> 25.60 | Improved |
| add adaptive sampling budget K (ablation) | MASS-S · SNU-FILM Extreme | PSNR: 25.60 -> 25.65; FLOPs 0.54 -> 0.49 | Improved |
| add velocity-aware step size (VA-SSM) (ablation) | MASS-S · SNU-FILM Extreme | PSNR: 25.65 -> 25.73 | Improved |
| use VA-SSM instead of 1D conv, LSTM or self-attention as integrator (ablation) | MASS-S · Xiph-4K | PSNR: 34.07/34.10/34.20 -> 34.65 | Improved |
| increase scale iterations from 1 to 2 (ablation) | MASS-S · Xiph-4K | PSNR: 34.40 -> 34.65 | Improved |
| increase scale iterations from 2 to 3 (ablation) | MASS-S · SNU-FILM Extreme | PSNR: 25.73 -> 25.75; FLOPs 0.50 -> 0.68 | No clear effect |
| add fusion head (softmax gating) for bidirectional contexts (ablation) | MASS-S · SNU-FILM Extreme | PSNR: 25.50 -> 25.73 | Improved |
| add refinement head using forward/backward discrepancy (ablation) | MASS-S · SNU-FILM Extreme | PSNR: 25.44 -> 25.73 | Improved |
| rotate trajectory direction 90 degrees from the flow (ablation) | MASS-S · Xiph-4K | PSNR: 34.65 -> 34.34 | Hurt |
| replace trajectory scan with static 3x3 grid order (ablation) | MASS-S · SNU-FILM Hard | PSNR: 30.90 -> 26.75 | Hurt |
| randomly permute or reverse the token order (ablation) | MASS-S · Xiph-2K | PSNR: 36.84 -> 34.86 (random), 33.44 (reverse) | Hurt |
| enable residual-velocity integration (delta v) (ablation) | MASS-S · Xiph-2K | PSNR: 36.76 -> 36.84 | 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.
- VFIMamba — VFIMamba: Video Frame Interpolation with State Space Models (NeurIPS 2024) · 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
- Sparse Global Matching for Video Frame Interpolation with Large Motion (CVPR 2024 · similarity 0.88)
- LC-Mamba: Local and Continuous Mamba with Shifted Windows for Frame Interpolation (CVPR 2025 · similarity 0.87)
- Generalizable Implicit Motion Modeling for Video Frame Interpolation (NeurIPS 2024 · similarity 0.87)
- EDEN: Enhanced Diffusion for High-quality Large-motion Video Frame Interpolation (CVPR 2025 · similarity 0.86)
- VFIMamba: Video Frame Interpolation with State Space Models (NeurIPS 2024 · similarity 0.86)
- BiM-VFI: Bidirectional Motion Field-Guided Frame Interpolation for Video with Non-uniform Motions (CVPR 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.
