readme_en.md

A webui for Propainter

A webui that you can easily pick up objects from the video and eliminate them.

Demo

Getting started

If you don't want to install the environment, you can download the package directly.
link 百度网盘
tutorial bilibili youtube

clone repo

git clone https://github.com/halfzm/ProPainiter-Webui.git

create conda enviroment

conda create -n propainter python=3.10
conda activate propainter

install dependencies

Just follow the instructions in Segment-ant-Track-Anything 和 ProPainter（P.S.Don't need to install groundingdino, I have put it in the project.）

pip install -r requirements.txt

prepare pretrained models

Download all the needed models for propainter
propainter
raft-things
recurrent_flow_completion
i3d_rgb_imagenet

Download all the needed models for segment-and-track-anything
SAM-VIT-B (sam_vit_b_01ec64.pth)
R50-DeAOT-L (R50_DeAOTL_PRE_YTB_DAV.pth)
GroundingDINO-T (groundingdino_swint_ogc)\

The directory structure will be arranged as：

ckpt
   |- bert-base-uncased
   |- groundingdino_swint_ogc.pth
   |- R50_EdAOTL_PRE_YTB_DAV.pth
   |- sam_vit_b_01ec64.pth
...
ProPainter/weights
   |- ProPainter.pth
   |- recurrent_flow_completion.pth
   |- raft-things.pth
   |- i3d_rgb_imagenet.pt (for evaluating VFID metric)
   |- README.md

quick start

python app.py

Reference

• Segment-ant-Track-Anything
• ProPainter

Star History