U^2-Net HTTP

This is an HTTP service wrapper for U^2-Net: Going Deeper with Nested U-Structure for Salient Object Detection.

This repository was inspired by cyrildiagne's basnet-http repository and referenced it.

Spedical thanks to @cyrildiagne !

Run U^2-Net HTTP server

python main.py

Post an image file

curl -F "[email protected]" http://localhost:8080 -o output.png

---

Development

1. Setup

This repository requires folliwing libraries.

• python 3.6.12
• pytorch 1.7.1
• torchvision 0.8.2
• numpy 1.17.*
• pillow 7.2.0
• scikit-image 0.17.2
• opencv-python 3.4.*
• gdown 3.12.2
• Flask 1.1.1
• flask-cors 3.0.8
• gunicorn 19.9.0

Note: If you have conda, you can create conda environment by following steps.

Example for CPU (Without GPU(CUDA))

conda create -n u2net-http python=3.6
activate u2net-http
conda install pytorch torchvision cpuonly -c pytorch
pip install -r requirements.txt

Note: If you can not call activate, try conda activate.

Of course, this repository will also work on GPU and Mac/Linux environments. For more information on how to install PyTorch, please refer to PyTorch official site.

2. Clone repositories

• Clone this U-2-Net-http repository

  git clone https://github.com/adakoda/U-2-Net-http.git
  

• Move to cloned directory and clone additional U^2-Net repository

  cd U-2-Net-http
  

  git clone https://github.com/adakoda/U-2-Net.git
  

3. Download pretrained model weight files

• Run model weights download script in U-2-Net-http/U-2-Net directory

  python setup_model_weights.py
  

  After finished python script, you will get these files.

  U-2-Net-http/U-2-Net/saved_models/u2net/u2net.pth 
  U-2-Net-http/U-2-Net/saved_models/u2net_portrait/u2net_portrait.pth