Dataset for image and video resolution enhancement assessment. The VPAD dataset consists of a set of videos for various movies, documentary, sports, and TV news channels with the presence of moving and text-like regions. The video sequences were obtained from a wide range of open websites and the Figure below shows a sample frame from each category. The dataset includes a total of 233 video sequences and is publicly released to encourage further research into projector resolution enhancement assessment in practical environments. More specifically, this dataset includes the following ten categories: Action, Comedy and Romance, Documentary, Fantasy, Graphics and Animation, Horror, News, Sports, TV Shows, and TV Episodes. The videos from the same category share some common features, such as similar background or contents.
The dataset, can be downloaded by clicking here (1.64GB), includes the following 10 categories.
Action Movies (15 videos, 92.1 MB)
Comedy/Romance Movies (22 videos, 119.7 MB)
Documentary (13 videos, 61.8 MB)
Fantasy Movies (62 videos, 484.8 MB)
Graphics or Animation (68 videos, 727.1 MB)
Horror Movies (4 videos, 40.4 MB)
News (11 videos, 15.3 MB)
Sports (8 videos, 25.6 MB)
TV Shows (13 videos, 8 MB)
TV Episodes (10 videos, 69.2 MB)
Related research areas
Please cite the following paper if you used this dataset:
- Xiaodan Hu, Mohamed A. Naiel, Alexander Wong, Mark Lamm, Paul Fieguth, "RUNet: A Robust UNet Architecture for Image Super-Resolution, in Press Computer Vision and Pattern Recognition Workshop, 2019. Details