Paper-BMVC | Download the paper [here]
The Dataset can be downloaded from this link
Download the model checkpoints (all stages) from this link
- Extract the model checkpoints to the root directory.
- Place the frames for videos in folders in the following format and pass the path to the folder containing videos (split into frames) to
root
.
root
└───videos_to_test
│ └───Video_1
| | │ │ xxxx.png
| | │ │ xxxx.png
| | │ └───...
│ └───Video_2
| | │ │ xxxx.png
| | │ │ xxxx.png
| | │ └───...
│ └───Video_3
| | │ │ xxxx.png
| | │ │ xxxx.png
| | │ └───...
- Modify the
test_only.py
according to the stage that you would like to test. (both checkpoint name and the function name to load the models). - The folder for output will be generated according to the checkpoint name.
If you find this useful for your research, please cite the following paper:
@article{ali2020deep,
title={Deep Motion Blind Video Stabilization},
author={Ali, Muhammad Kashif and Yu, Sangjoon and Kim, Tae Hyun},
journal={arXiv preprint arXiv:2011.09697},
year={2020}
}