Background Removal API Using Deep Learning Tool

• Here i am creating a FastAPI for removes the Background of an Images using Python And pre-trained AI/ML models which accepts images with .jpg and .png extension.

Steps That I follow to develope this task.

• There is a one python library or we can say a project or a tool that can help us to remove background of an image files

• https://github.com/danielgatis/rembg

• It uses the u2net: salienct object detection(SOD) a deep learning model

Download the model (Not Required - rembg automatically download it for us)

• https://github.com/danielgatis/rembg/releases/download/v0.0.0/u2net.onnx
• https://github.com/danielgatis/rembg/releases/download/v0.0.0/u2netp.onnx

How to run the FastAPI

1. Create a python virtual environment and activate it:

   • python3 -m venv venv
   • source venv/bin/activate

2. install all the required libraries using requirements.txt

   • pip install -r requirements.txt

3. run the API

   • python3 app.py

4. open url in the browser

   • http://127.0.0.1:8000/docs

5. upload the image (.jpg, .jpeg , .png) in FastAPI Swagger UI

6. we get the output as dict with base64 of the image

For the demo purpose I create a Streamlit App

How to run the Streamlit app

• go to the path where streamlit_app_demo.py is located and run the below command
• streamlit run streamlit_app_demo.py

Other References

• Official repository of U-2-Net
  • https://github.com/xuebinqin/U-2-Net

Reserch paper for the u2net

• U2-Net: Going Deeper with Nested U-Structure for Salient Object Detection
• https://arxiv.org/pdf/2005.09007.pdf