UPDATE 12/2024: A large-scale high resolution dataset for video interpolation is now available check out LAVIB

Inter4K Tools

Official repository for hosting tools for creating Inter4K video interpolation dataset.

Project website: Inter4K/index.html

Paper: AdaPool: Exponential Adaptive Pooling for Information-Retaining Downsampling

Requirements

• FFmpeg (https://www.ffmpeg.org/download.html)

Dataset creation instructions

1. You will first need to download the 4K 60fps video clips (~15GB in total). Use the link https://tinyurl.com/inter4KUHD either through the browser or with wget on terminal:

wget https://tinyurl.com/inter4KUHD

2. Clone the repository:

git clone git@github.com:alexandrosstergiou/Inter4K.git

3. Unzip the folder inside the repository:

unzip Inter4K.zip -d Inter4K

4. Perform a check to all files by running:

cd Inter4K/
python check.py

5. Finally, create the full dataset by running:

python resolution_changes.py

! Note: By default, the program will use 16 workers to create the dataset. If you are interested on speeding-up things (and given that you have a capable CPU) you can change the number of workers by editing resolution_chanes.py line #42. This can also be the case that a less powerful CPU is used.

Reference

@article{stergiou2021adapool,
  title={AdaPool: Exponential Adaptive Pooling for Information-Retaining Downsampling},
  author={Stergiou, Alexandros and Poppe, Ronald},
  booktitle={arXiv},
  year={2021}
}