Getting started with Docker

In order to replicate the results smoothly and avoid dependency errors (aka CUDA installation hell) you can use Docker combined with NVIDIA-Docker. Docker will install all the packages in an isolated environment.

Note: You will need an NVIDIA GPU and a Linux OS to use NVIDIA-Docker.

Installing Docker

• Docker
• Docker-Compose
• nvidia-docker

Downloading the required files

• Download the frames and masks folders from here and place them on the demo folder.

• Download the files FlowNet2_checkpoint.pth.tar, imagenet_deepfill.pth and resnet101_movie.pth from here and place them in pretrained_models.

├── demo
│   ├── frames
│   └── masks
├── pretrained_models
│   ├── FlowNet2_checkpoint.pth.tar
│   ├── imagenet_deepfill.pth
│   └── resnet101_movie.pth

Usage

1. From the docker folder run: docker-compose up -d

2. Access the conatiner: docker exec -it inpainting bash

That will open a CLI on the Docker container. Now you can run the demo scripts, for example:

python3 tools/video_inpaint.py --frame_dir ./demo/frames --MASK_ROOT ./demo/masks --img_size 512 832 --FlowNet2 --DFC --ResNet101 --Propagation

Tested on Ubuntu 18.04 with a GTX 1060 GPU (drivers 410.104). Not working on higher architectures such as sm_75 (Turing), e.g. RTX 2080 Ti