- B06607057
- R10944012
- R10944014
- Ubuntu 18.04
- Nvidia RTX 3090, Driver Version 510.73.08, cuda 11.6
- Docker Engine 20.10.14
- Docker Compose 1.29.2
- Install
dockeranddocker-compose - Install Nvidia Drivers and Setup Cuda Environment
Note: if no cuda available, or not supported, please comment out the following line in the
docker-compose.yml
`runtime: nvidia` 3. cd to `code` directory and extract backend.zip ``` bash cd code unzip backend.zip ``` > Note: Source code for frontend is available at `frontend.zip`, but the built version is included in backend. 4. Put the Input Video in the `backend/static/videos` folder and Rename the Video to `video.mp4` >Note: We assume the video resolution is smaller or equals to 1280x720, and has fps 30. In addition, the ebsynth algorithm is pretty slow, a 4-second video can take about 30 minutes to compute. One can use the following command to change the resolution, fps and cut the video into the desired length:
`ffmpeg -i -ss -to -vf "fps=30, scale=-1:720" video.mp4` 4. Run `docker-compose up` > Note: The docker-compose will build the docker image and start the backend server. 5. The frontend will be at `localhost:12113` > Note: The frontend cannot access by multiple client, or it may cause errors. 6. The styled video will be at `backend/static/videos/video_styled.mp4`
This System is based on the following libraries and pretrained model: