Video Super-Resolution with Convolutional Neural Networks
Event details
| Date | 26.08.2016 |
| Hour | 10:00 › 12:00 |
| Speaker | Ruofan Zhou |
| Location | |
| Category | Conferences - Seminars |
EDIC Candidacy Exam
Exam President: Prof. Pascal Fua
Thesis Director: Prof. Sabine Süsstrunk
Co-examiner: Prof. Wenzel Jakob
Background Papers
Large-scale Video Classification with Convolutional Neural Networks by Andrej Karpathy, George Toderici, Sanketh Shetty, Thomas Leung, Rahul Sukthankar, Li Fei-Fei
Learning a Deep Convolutional Network for Image Super-Resolution by Chao Dong, Chen Change Loy, Kaiming He, Xiaoou Tang
On Bayesian Adaptive Video Super Resolution by Ce Liu, Deqing Sun
Abstract;
Super resolving real-world video sequences remains to be a challenging research task in computer vision field. While nowadays convolutional neural networks have achieved state-of-the-art performance on many vision tasks. In this proposal, we discuss three existing works on super-resolution task or CNN architecture that inspire us on applying CNN on video super-resolution task. The first paper has proposed a Bayesian-based model for video super-resolution task that estimate motion, blur kernel and noise level simultaneously to adapt to a variety of noise levels and blur kernels. Using different convolutional neural networks, the later two papers achieve promising results on single-image super resolution and video classification respectively.
Based on the key ideas from these three papers, we propose our research plan to design a CNN architecture for video super-resolution.
Exam President: Prof. Pascal Fua
Thesis Director: Prof. Sabine Süsstrunk
Co-examiner: Prof. Wenzel Jakob
Background Papers
Large-scale Video Classification with Convolutional Neural Networks by Andrej Karpathy, George Toderici, Sanketh Shetty, Thomas Leung, Rahul Sukthankar, Li Fei-Fei
Learning a Deep Convolutional Network for Image Super-Resolution by Chao Dong, Chen Change Loy, Kaiming He, Xiaoou Tang
On Bayesian Adaptive Video Super Resolution by Ce Liu, Deqing Sun
Abstract;
Super resolving real-world video sequences remains to be a challenging research task in computer vision field. While nowadays convolutional neural networks have achieved state-of-the-art performance on many vision tasks. In this proposal, we discuss three existing works on super-resolution task or CNN architecture that inspire us on applying CNN on video super-resolution task. The first paper has proposed a Bayesian-based model for video super-resolution task that estimate motion, blur kernel and noise level simultaneously to adapt to a variety of noise levels and blur kernels. Using different convolutional neural networks, the later two papers achieve promising results on single-image super resolution and video classification respectively.
Based on the key ideas from these three papers, we propose our research plan to design a CNN architecture for video super-resolution.
Practical information
- General public
- Free
Contact
- Cecilia Chapuis