Research map / Video / Video segmentation and tracking
Multi-Temporal Recurrent Neural Networks For Progressive Non-Uniform Single Image Deblurring With Incremental Temporal Training
Dongwon Park, Dong Un Kang, Jisoo Kim, Se Young Chun
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 kernel size 3 instead of 5 (ablation) | MS network of Tao · GoPro (ablation, intermediate-TL dataset) | PSNR (dB): 29.97 -> 30.10 | Improved |
| add residual learning (ablation) | MS network · GoPro (ablation, intermediate-TL dataset) | PSNR (dB): 30.10 -> 30.25 | Improved |
| add residual skip connection linked with I0 (ablation) | MT-RNN · GoPro with intermediate TL images | PSNR (dB): 0.15dB | Improved |
| replace multi-scale with multi-temporal approach (ablation) | Tao-style network, K=3, residual · GoPro (ablation, intermediate-TL dataset) | PSNR (dB): 30.25 -> 30.43 | Improved |
| add recurrent feature maps (ablation) | MT-RNN · GoPro (ablation, intermediate-TL dataset) | PSNR (dB): 30.43 -> 30.74 | Improved |
| combine MS with MT instead of MT alone (ablation) | MT-RNN · GoPro (ablation, intermediate-TL dataset) | PSNR (dB): 30.74 -> 30.58 | Hurt |
| use 2 temporal steps instead of one-stage (0 steps) (ablation) | MT-RNN · GoPro (temporal steps and size ablation) | PSNR (dB): 29.93 -> 30.74 | Improved |
| use temporal step 4 instead of 2 (ablation) | MT-RNN · GoPro (temporal steps and size ablation) | PSNR (dB): 30.74 -> 30.44 | Hurt |
| halve parameter size (1.46M vs 2.63M) (ablation) | MT-RNN · GoPro (temporal steps and size ablation) | PSNR (dB): 30.74 -> 30.21 | Hurt |
| double parameter size (5.35M vs 2.63M) (ablation) | MT-RNN · GoPro (temporal steps and size ablation) | PSNR (dB): 30.74 -> 30.84 | No clear effect |
| apply MT approach with incremental temporal training (ablation) | Zhang network · GoPro | PSNR (dB): 30.25 -> 30.91 | Improved |
| apply MT approach with incremental temporal training (ablation) | Kupyn network · GoPro | PSNR (dB): 28.27 -> 28.36 | No clear effect |
Closest papers by meaning
- Scale-Recurrent Network for Deep Image Deblurring (CVPR 2018 · similarity 0.91)
- Deep Multi-Scale Convolutional Neural Network for Dynamic Scene Deblurring (CVPR 2017 · similarity 0.89)
- Efficient Multi-scale Network with Learnable Discrete Wavelet Transform for Blind Motion Deblurring (CVPR 2024 · similarity 0.89)
- Online Video Deblurring via Dynamic Temporal Blending Network (ICCV 2017 · similarity 0.88)
- Deblurring Dynamic Scenes via Spatially Varying Recurrent Neural Networks (TPAMI 2021 · similarity 0.88)
- Learning Scale-Aware Spatio-temporal Implicit Representation for Event-based Motion Deblurring (ICML 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.
