/video-object-removal

Just draw a bounding box and you can remove the object you want to remove.

Primary LanguagePythonMIT LicenseMIT

video-object-removal

Just draw a bounding box and you can remove the object you want to remove.

Installation

All the code has been tested on Ubuntu 16.04, Python 3.5, Pytorch 0.4.0, CUDA 8.0, GTX1080Ti GPU.

  • Clone the repository
git clone https://github.com/zllrunning/video-object-removal.git
cd video-object-removal
cd get_mask
bash make.sh
cd ../inpainting
bash install.sh
cd ..

Demo

  • Download pretrained models of SiamMask and Inpainting
  • Put them in cp/ folder
  • Then just run:
python demo.py --data data/Human6
  • It also supports video file.
python demo.py --data data/bag.avi
  • Another optional parameter : --mask-dilation
python demo.py --data data/Human6  --mask-dilation 24

This parameter controls the size of the dilation kernel used for the mask. The role is to expand the range of the mask to avoid edge problems. Please see inpainting/davis.py for more details.


1. Just draw a bounding box like this:

2. The objected will be removed and the inpainted video will be saved in results/inpainting folder. (The Gif image loading takes some time, please wait a moment.)

Examples

Acknowledgement

Citation

@article{Wang2019SiamMask,
    title={Fast Online Object Tracking and Segmentation: A Unifying Approach},
    author={Wang, Qiang and Zhang, Li and Bertinetto, Luca and Hu, Weiming and Torr, Philip HS},
    journal={The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
    year={2019}
}
@inproceedings{kim2019deep,
  title={Deep Video Inpainting},
  author={Kim, Dahun and Woo, Sanghyun and Lee, Joon-Young and So Kweon, In},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  pages={5792--5801},
  year={2019}