Z-MU-Z / Click2Trimap

Repository from Github https://github.comZ-MU-Z/Click2TrimapRepository from Github https://github.comZ-MU-Z/Click2Trimap

Learning Trimaps via Clicks for Image Matting

Chenyi Zhang, Yihan Hu, Henghui Ding, Humphrey Shi, Yao Zhao, Yunchao Wei

drawing

Quick Installation

Our demo is developed in : Python3.8, PyTorch 1.10.1, CUDA 11.3.

Run the following command to install required packages. You can refer to the issues of RITM, SimpleClick and AEMatter if you run into problems.

pip3 install -r requirements.txt

Demo

drawing

drawing

drawing

A command to run the demo.

python3 demo.py --checkpoint-c2t='path to click2trimap checkpoint' --checkpoint-matting='path to matting model checkpoint'

Images can be found here.

Guidelines of Clicking

drawing

Tip 1:
Click the left mouse button on foreground region.
Click the roll mouse button on unknown region.
Click the right mouse button on background region.

Tip 2:
Refer to the gif for how to provide appropriate clicks.

Tip 3:
Using images with high resolutions leads to longer waiting time and might even run out of memory.

Download

Checkpoints:

Click2Trimap: Google Drive

Matting Model: AEMatter

We choose AEMatter as our trimap-based matting model in this demo because of its excellent performance and friendly memory usage. But it's also convenient to switch to any other trimap-based matting model if you want.

To do list

  • Web UI
  • HuggingFace Demo
  • ...

License

The code is released under the MIT License. It is a short, permissive software license. Basically, you can do whatever you want as long as you include the original copyright and license notice in any copy of the software/source.

Citation

@misc{zhang2024learning,
      title={Learning Trimaps via Clicks for Image Matting}, 
      author={Chenyi Zhang and Yihan Hu and Henghui Ding and Humphrey Shi and Yao Zhao and Yunchao Wei},
      year={2024},
      eprint={2404.00335},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Acknowledgement

Our project is developed based on RITM, SimpleClick and AEMatter. Thanks for their nice demo GUI and wonderful work!

About

License:MIT License


Languages

Language:Python 99.7%Language:Cython 0.3%