Research map / Video / Video segmentation and tracking
Motion-Aware Dynamic Architecture for Efficient Frame Interpolation
Myungsub Choi, Suyoung Lee, Heewon Kim, Kyoung Mu Lee
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 the dynamic scale and depth framework (CAIN-SD) to CAIN | CAIN · Xiph-2K | TFLOPs: 3.133 -> 1.598 | Improved |
| add the dynamic scale and depth framework (CAIN-SD) to CAIN | CAIN · Xiph-2K | PSNR (dB): 35.21 -> 34.68 | Hurt |
| add the dynamic scale and depth framework (CAIN-SD) to CAIN | CAIN · Xiph-4K | PSNR (dB): 32.56 -> 32.92 | Improved |
| add the dynamic scale and depth framework (CAIN-SD) to CAIN | CAIN · Xiph-2K | GPU time (s): 0.225 -> 0.237 | Hurt |
| add the dynamic framework (CAIN-SD) to CAIN | CAIN · Xiph-2K | CPU time (s): 10.64 -> 8.83 | Improved |
| add the depth finder to scale-only model (CAIN-S to CAIN-SD) (ablation) | CAIN-S · REDS-120fps val | TFLOPs: 1.507 -> 0.882 | Improved |
| use scale-only variant (CAIN-S) instead of CAIN (ablation) | CAIN · REDS-120fps val | TFLOPs: 1.305 -> 1.507 | Hurt |
| add scale selection to depth-only model (CAIN-D to CAIN-SD) (ablation) | CAIN-D · REDS-120fps val | TFLOPs: 1.008 -> 0.882 | Improved |
| lower depth weight lambda_d from 5e-9-ratio setting to 5e-10 with lambda_s fixed (ablation) | CAIN-SD · Xiph-2K | TFLOPs: 1.598 -> 3.804; PSNR 34.68 -> 35.33 | Hurt |
| increase lambda_s to 5e-8 (ablation) | CAIN-SD · Xiph-2K | PSNR (dB): 34.68 -> 31.10; TFLOPs 1.598 -> 1.827 | Hurt |
| decrease lambda_s to 5e-10 (ablation) | CAIN-SD · Xiph-2K | PSNR (dB): 34.68 -> 34.71; TFLOPs 1.598 -> 1.699 | Mixed |
| test with full-frame input instead of 256 patches (ablation) | CAIN-SD · Xiph-2K | PSNR (dB): 34.68 -> 33.24 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- AdaCoF+ — AdaCoF: Adaptive Collaboration of Flows for Video Frame Interpolation (CVPR 2020) · better on 0 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
- SportsSloMo: A New Benchmark and Baselines for Human-centric Video Frame Interpolation (CVPR 2024 · similarity 0.86)
- EDEN: Enhanced Diffusion for High-quality Large-motion Video Frame Interpolation (CVPR 2025 · similarity 0.85)
- Following Motion for Sequential Modeling in Video Frame Interpolation (ECCV 2026 · similarity 0.85)
- Aligning Anything: Hierarchical Motion Estimation for Video Frame Interpolation (ECCV 2026 · similarity 0.85)
- Framer: Interactive Frame Interpolation (ICLR 2025 · similarity 0.84)
- Generative Inbetweening: Adapting Image-to-Video Models for Keyframe Interpolation (ICLR 2025 · similarity 0.82)
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.
