Research map / Video / Video segmentation and tracking
BiFormer: Learning Bilateral Motion Estimation via Bilateral Transformer for 4K Video Frame Interpolation
Junheum Park, Jintae Kim, Chang-Su Kim
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 BiFormer instead of the strongest listed baseline, M2M-PWC | 4K video frame interpolation · X4K1000FPS | PSNR: 30.81 -> 31.32 | Improved |
| Use BiFormer instead of the strongest listed baseline, M2M-PWC | 4K video frame interpolation · Xiph-4K | PSNR: 34.46 -> 34.48 | Improved |
| Use BiFormer instead of the strongest listed baseline, M2M-PWC | 4K video frame interpolation · BVI-DVC-4K | PSNR: 29.77 -> 29.67 | Hurt |
| Add bilateral cross-attention blocks to the six-Swin-block estimator, including BCA-A and two BCA+A blocks (ablation) | BiFormer global motion estimator · X4K1000FPS | PSNR: 30.94 -> 31.32 | Improved |
| Add bilateral cross-attention blocks to the six-Swin-block estimator, including BCA-A and two BCA+A blocks (ablation) | BiFormer global motion estimator · X4K1000FPS | SSIM: 0.8867 -> 0.9212 | Improved |
| Use BBCV search range r=2 rather than r=1 (ablation) | BiFormer local motion refinement · X4K1000FPS | PSNR: 31.20 -> 31.32 | Improved |
| Use BBCV search range r=2 rather than r=1 (ablation) | BiFormer local motion refinement · X4K1000FPS | SSIM: 0.9037 -> 0.9212 | Improved |
| Include larger BBCV blocks, up to 4×4, rather than limiting the maximum to 2×2 (ablation) | BiFormer local motion refinement · X4K1000FPS | PSNR: 31.27 -> 31.32 | Improved |
| Include larger BBCV blocks, up to 4×4, rather than limiting the maximum to 1×1 (ablation) | BiFormer local motion refinement · X4K1000FPS | PSNR: 31.16 -> 31.32 | Improved |
| Refine motion sequentially at 1/4 and 1/2 scales rather than at 1/2 scale only (ablation) | BiFormer local motion refinement · X4K1000FPS | PSNR: 31.15 -> 31.32 | Improved |
| Refine motion sequentially at 1/4 and 1/2 scales rather than at 1/4 scale only (ablation) | BiFormer local motion refinement · X4K1000FPS | PSNR: 30.88 -> 31.32 | Improved |
Later papers that compared with this method
1 later accepted paper reports BiFormer as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Sparse Global Matching for Video Frame Interpolation with Large Motion (CVPR 2024 · better on 5 of 8 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
- Asymmetric Bilateral Motion Estimation for Video Frame Interpolation (ICCV 2021 · similarity 0.89)
- BMBC: Bilateral Motion Estimation with Bilateral Cost Volume for Video Interpolation (ECCV 2020 · similarity 0.89)
- BiM-VFI: Bidirectional Motion Field-Guided Frame Interpolation for Video with Non-uniform Motions (CVPR 2025 · similarity 0.86)
- Super SloMo: High Quality Estimation of Multiple Intermediate Frames for Video Interpolation (CVPR 2018 · similarity 0.85)
- MBTI: Masked Blending Transformers with Implicit Positional Encoding for Frame-rate Agnostic Motion Estimation (ICCV 2025 · similarity 0.85)
- AMT: All-Pairs Multi-Field Transforms for Efficient Frame Interpolation (CVPR 2023 · similarity 0.84)
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.
