Invisibility Cloak Project
Computer Vision with OpenCV
Graduation Project in 2021 Fall Semester @UNIST
For this project, FGVC(Flow-edge Guided Video Completion) deep learning model was used for object removal.
| Package | Version |
|---|---|
| anaconda (x64) | 4.10.3 |
| cuda | 10.2.89 |
| matplotlib | 3.4.3 |
| numpy | 1.21.4 |
| opencv | 4.5.4 |
| os | Windows 10 |
| pip | 21.0.1 |
| python | 3.8.12 |
| pytorch | 1.6.0 |
| scipy | 1.6.2 |
$ conda create -n (name)
$ conda activate (name)
$ conda install pytorch==1.6.0 torchvision==0.7.0 cudatoolkit=10.2 -c pytorch
$ pip install matplotlib scipy opencv imageio imageio-ffmpeg scikit-image imutils- Download and unzip weight.zip into the
modules. - Prepare video sequences dataset of color and mask for project. (Data Samples : tennis | f250)
- Run project
# Remove __pycache__ if you want.
$ find . | grep -E "(__pycache__|\.pyc|\.pyo$)" | xargs rm -rf
# Run video inpainting.
$ python run_inpainting.py \
> --path './data/tennis_color' \
> --path_mask './data/tennis_mask' \
> --outroot './data/tennis_result' \
> --merge \
> --run- Add to remove object with FGVC
- Set the main cloak color with RED
- Update cloak mask and noise.
- Rename ftn -> image_stack.
- Add to save output video.
- Add color selection mode : RED || GREEN
- Get object removal result by FGVC.
- Add specific noise filtering conditions.
- Modify detected color : RED -> GREEN
- Project first commit.
- Test module ftn for showing image stack view.
- Test HSV Detector using track bar.
- Test Invisibility Cloak Demo.
Updated : 2021-11-28 00:19