Research map / Video / Video segmentation and tracking
Gyro-based Deep Video Deblurring
Jaesung Rim, Woohyeok Kim, Haeyun Lee, Heemin Yang, Ke Wang, Sunghyun Cho
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 rotational blur kernels from gyro (offsets and shift patterns) (ablation) | GyroDVD-64 · GyroVD-Syn (150K-iter ablation, GyroDVD-64) | PSNR (dB): 33.51 -> 34.30 | Improved |
| add translational kernel from optical flow on top of rotational kernels (ablation) | GyroDVD-64 · GyroVD-Syn (150K-iter ablation, GyroDVD-64) | PSNR (dB): 34.30 -> 34.65 | Improved |
| use video frame to estimate deformable offsets instead of blur kernels (ablation) | GyroDVD-64 · GyroVD-Syn (150K-iter ablation, GyroDVD-64) | PSNR (dB): 33.51 -> 33.51 | No clear effect |
| use optical flows instead of blur kernels for offsets (ablation) | GyroDVD-64 · GyroVD-Syn (150K-iter ablation, GyroDVD-64) | PSNR (dB): 33.51 -> 34.11 | Improved |
| compute ktran from GT sharp frames instead of blurred frames (ablation) | GyroDVD-64 · GyroVD-Syn (150K-iter ablation, GyroDVD-64) | PSNR (dB): 34.65 -> 34.78 | No clear effect |
| add blur kernels via concatenation in the video decoder (ablation) | GyroDVD-64 · GyroVD-Syn (150K-iter ablation, GyroDVD-64) | PSNR (dB): 33.95 -> 33.42 | Hurt |
| add KGC in the video decoder (ablation) | GyroDVD-64 · GyroVD-Syn (150K-iter ablation, GyroDVD-64) | PSNR (dB): 33.95 -> 33.62 | Hurt |
| replace fixed shift with Kernel-Guided Shift Block in decoder (ablation) | GyroDVD-64 · GyroVD-Syn (150K-iter ablation, GyroDVD-64) | PSNR (dB): 33.95 -> 34.65 | Improved |
| use KAM in decoder instead of KGS-Block (ablation) | GyroDVD-64 · GyroVD-Syn (150K-iter ablation, GyroDVD-64) | PSNR (dB): 34.65 -> 34.15 | Hurt |
| fuse krot and ktran in feature domain instead of pixel domain (ablation) | GyroDVD-64 · GyroVD-Syn (150K-iter ablation, GyroDVD-64) | PSNR (dB): 34.55 -> 34.65 | Improved |
| add gyro-based kernels to video deblurring (ShiftNet with krot vs ShiftNet) | ShiftNet · GyroVD-Syn (average) | PSNR (dB): 34.37 -> 34.55 | Improved |
| use GyroDVD-128 instead of ShiftNet+ | GyroDVD · GyroVD-Syn (average) | PSNR (dB): 35.31 -> 35.93 | Improved |
Closest papers by meaning
- Gyro-based Neural Single Image Deblurring (CVPR 2025 · similarity 0.94)
- DynaMoDe-NeRF: Motion-aware Deblurring Neural Radiance Field for Dynamic Scenes (CVPR 2025 · similarity 0.86)
- GS-Blur: A 3D Scene-Based Dataset for Realistic Image Deblurring (NeurIPS 2024 · similarity 0.86)
- Spike-guided Motion Deblurring with Unknown Modal Spatiotemporal Alignment (CVPR 2024 · similarity 0.85)
- Real-World Efficient Blind Motion Deblurring via Blur Pixel Discretization (CVPR 2024 · similarity 0.85)
- Parameterized Blur Kernel Prior Learning for Local Motion Deblurring (CVPR 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.
