/AdaMatting

an unofficial pytorch implementation of ICCV 2019 paper "Disentangled Image Matting"

Primary LanguagePythonMIT LicenseMIT

AdaMatting

an unofficial pytorch implementation of ICCV 2019 paper "Disentangled Image Matting"
this project is not finished

Dependencies

  • python 3.6
  • pytorch 1.4.0
  • tensorboardX
  • opencv

Datasets

Adobe Composition-1k Dataset

Contact author for the dataset.

MSCOCO

train2014.zip

PASCAL VOC

PASCAL VOC:

Usage

Download and move all above compressed data files into a folder.
Set --raw_data_path={path_to_your_folder} in any .sh file you use.

Decompress and composite the data

$ bash ./prep.sh

Train

$ bash ./train.sh

Visualize training

$ tensorboard --logdir ./runs

Test

TODO

Results and trained model

TODO