AFD-former: A Hybrid Transformer with Asymmetric Flow Division for Synthesized View Quality Enhancement
Xu Zhang1, Nian Cai1, Huan Zhang1, Yun Zhang2, Jianglei Di1, Weisi Lin3,
1Guangdong University of Technology, 2Sun Yat-sen University, 3Nanyang Technological University.
- Datasets
- Evaluation
| Setting | Dataset Name | Index |
|---|---|---|
| NO. 1 | Balloons | l1~l5:1-20 |
| NO. 2 | Bookarrival | l1~l5:21-30 |
| NO. 3 | Undodancer | l1~l5:31-50 |
| NO. 4 | Gtfly | l1~l5:51-70 |
| NO. 5 | Outdoor | l1~l5:71-80 |
| NO. 6 | Poznancarpark | l1~l5:81-100 |
| NO. 7 | PoznanStreet | l1~l5:101-120 |
| NO. 8 | Shark | l1~l5:121-140 |
Note:
- Patch-image includes training and validation datasets.
- l1~l5 represent the distortion levels from 1 to 5, ordered by severity.
| Types | Datasets |
|---|---|
| H.264 | newspaper |
| H.264 | poznanHall2 |
| H.264 | kendo |
| H.264 | lovebird1 |
| H.265 | pantomime |
| H.265 | newspaper |
| H.265 | poznanHall2 |
| H.265 | kendo |
| H.265 | lovebird1 |
| MCL-3D | MCL-3D |
| IETR | IETR |
If you find this work useful for your research, please consider citing:
@ARTICLE{AFD-former,
author={Zhang, Xu and Cai, Nian and Zhang, Huan and Zhang, Yun and Di, Jianglei and Lin, Weisi},
journal={IEEE Transactions on Circuits and Systems for Video Technology},
title={AFD-Former: A Hybrid Transformer With Asymmetric Flow Division for Synthesized View Quality Enhancement},
year={2023},
volume={33},
number={8},
pages={3786-3798},
doi={10.1109/TCSVT.2023.3241920}}
This project is mainly based on Restormer, NAFNet, and SRMNet. Thanks for their awesome works.
Please feel free to contact if there is any question (Xu Zhang: [email protected]).