Dataset preview

In this section you can observe the frames from the videos included in the dataset.

1 / 17

2 / 17

3 / 17

4 / 17

5 / 17

6 / 17

7 / 17

8 / 17

9 / 17

10 / 17

11 / 17

12 / 17

13 / 17

14 / 17

15 / 17

16 / 17

17 / 17

❮ ❯

Dataset description

The dataset of MSU VFI Benchmark consists of two parts: outdoor & indoor shooting and gaming content. Characteristics of first part:

• was recorded on iPhone 11 at 240 FPS
• 1920×1080 resolution
• 10 videos with various content and motion:
  • playing dogs
  • skating rink
  • indoor shooting with cat and items
  • textures

Characteristics of second part:

• was captured with OBS at 120 FPS
• 1920×1080 resolution
• 7 games including arcades, first-person shooters and racing

For dataset were chosen frame subsequences with 1 second length:

• 241 frames for videos recorded on camera
• 121 frames for games

Dataset categories

The dataset is divided into two categories: interpolation from 30 FPS and 60 FPS to the original FPS.

In 30 FPS category you need to interpolate:

• 7 frames between every pair of adjacent frames for 240 FPS videos
• 3 frames between every pair of adjacent frames for 120 FPS videos

In 60 FPS category you need to interpolate:

• 3 frames between every pair of adjacent frames for 240 FPS videos
• 1 frame between every pair of adjacent frames for 120 FPS videos

Dividing into these categories allows us to pay attention to the algorithms’ ability to cope with different time distances between frames.

Download

Low FPS version of the Dataset is available here.
Note that the original FPS version of Dataset have not been published and will not be published in the future.