Awesome
π DSTNet+
<!-- > [Jinshan Pan](https://jspan.github.io/), [Long Sun](https://github.com/sunny2109), [Boming Xu](https://github.com/xuboming8), [Jinhui Tang](https://scholar.google.com/citations?user=ByBLlEwAAAAJ&hl=zh-CN), and [Jiangxin Dong](https://scholar.google.com/citations?user=ruebFVEAAAAJ&hl=zh-CN&oi=ao) > [IMAG Lab](https://imag-njust.net/), Nanjing University of Science and Technology --><a href="https://colab.research.google.com/drive/19DdsNFeOYR8om8QCCi9WWzr_WkWTLHZd?usp=sharing"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="google colab logo"></a>
![]()
<br>
This repo is a official implementation of "Learning Efficient Deep Discriminative Spatial and Temporal Networks for Video Deblurring".
DSTNet+ is an extension of DSTNet.
Update
- 2023.01.08: This repo is created.
Results
-
Model efficiency (PSNR vs. Runtime vs. Params)
<img width="770" src="figs/runtime.png"> -
Quantitative evaluations <br> β β β β βEvaluation on GoPro dataset β β β β β β β Evaluation on DVD dataset <br> <img width="380" src="figs/table_gopro.png"> <img width="330" src="figs/table_dvd.png">
-
Deblurred results on GoPro dataset
<img width="780" src="figs/gopro.png"> -
Deblurred results on DVD dataset
<img width="780" src="figs/dvd.png"> -
Deblurred results on Real-world blurry frames
<img width="800" src="figs/real_world.png">