foamliu/Deep-Image-Matting
TensorFlow/Keras implementation of a deep neural network for image matting (foreground extraction from images).
Not currently ranked — collecting fresh signals.
star history
This repository reproduces the Deep Image Matting paper, implementing a deep learning model that separates foreground objects from backgrounds in images. It uses VGG16 as a backbone with a custom matting architecture, trained on the Adobe Deep Image Matting dataset along with MSCOCO and PASCAL VOC data. The project includes data preprocessing scripts, training code with TensorBoard logging, and a demo script for running inference with a pre-trained model.
Frequently asked
- What is foamliu/Deep-Image-Matting?
- TensorFlow/Keras implementation of a deep neural network for image matting (foreground extraction from images).
- Is Deep-Image-Matting open source?
- Yes — foamliu/Deep-Image-Matting is open source, released under the MIT license.
- What language is Deep-Image-Matting written in?
- foamliu/Deep-Image-Matting is primarily written in Python.
- How popular is Deep-Image-Matting?
- foamliu/Deep-Image-Matting has 991 stars on GitHub.
- Where can I find Deep-Image-Matting?
- foamliu/Deep-Image-Matting is on GitHub at https://github.com/foamliu/Deep-Image-Matting.