legendary-giggle: A Background Removal Tool with U-2-Net

Overview

This is a background removal tool based on the U-2-Net model, designed for efficient and accurate segmentation of objects within images. The tool takes an input image and outputs a version with the background removed, providing a clean and isolated foreground.

• U-2-Net Model: Utilizes the U-2-Net architecture for image segmentation.

Requirements

• Python 3.x
• Dependencies listed in requirements.txt

Installation

1. Clone the repository:

   git clone https://github.com/ShivamMadlani/legendary-giggle
   cd legendary-giggle

2. Install dependencies(if running on a local machine):

   pip install -r requirements.txt

3. Download u2net model:

   • Download the model from here
   • paste the downloaded file u2net.pth in saved_models/u2net

Usage

1. Run the background removal tool:

   python main.py

2. Add the images into the test_data/images/input folder.

3. The masked images will be generated inside test_data/images/u2net_results.

4. The isolated image will be generated in test_data/images/outpur folder.

Acknowledgments

• U-2-Net Model: Link to U-2-Net Repository