https://github.com/voinic/u2net-keras

Minimal scripts for testing U-2-Net models in Keras
https://github.com/voinic/u2net-keras

image-segmentation keras python-image-processing tensorflow u-2-net u2net

Last synced: 2 months ago
JSON representation

Minimal scripts for testing U-2-Net models in Keras

• Host: GitHub
• URL: https://github.com/voinic/u2net-keras
• Owner: Voinic
• License: mit
• Created: 2021-01-18T15:44:23.000Z (about 4 years ago)
• Default Branch: master
• Last Pushed: 2021-01-20T14:02:25.000Z (about 4 years ago)
• Last Synced: 2024-06-18T15:00:05.811Z (8 months ago)
• Topics: image-segmentation, keras, python-image-processing, tensorflow, u-2-net, u2net
• Language: Python
• Homepage:
• Size: 4.98 MB
• Stars: 6
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# u2net-keras

Simple scripts for testing U-2-Net models in Keras. 


Based on https://github.com/NathanUA/U-2-Net and https://github.com/shreyas-bk/U-2-Net-Keras 


Keras u2net, u2netp and u2net_portrait models were converted from the original pytorch models. 


# Quick start

1. Clone this repo to local

```

git clone https://github.com/Voinic/u2net-keras.git

```

2. Get models from [**here**](https://drive.google.com/drive/folders/1oLFscPEoqxHIcRvNT80ps4XAH9ab5LH2)

3. Test u2net/u2netp (background substraction) models

```

python u2net-keras.py images/input.jpg images/output.png

```

4. Test u2net_portrait model

```

python u2net_portrait-keras.py images/input_portrait.jpg images/output_portrait.png

```

# Results

Background substraction model 


![input](images/input.jpg) 


![output](images/output.png) 


Portrait creation model 


![input_portrait](images/input_portrait.jpg) 


![output_portrait](images/output_portrait.png) 


# Known issues

- Dont work with tensorflow without CUDA support.