BEN2 Background Remover is an AI-powered tool designed to remove backgrounds from images with high precision. It utilizes the BEN_Base model for processing images and is implemented using Gradio for an interactive web-based interface.

• Removes backgrounds from images efficiently
• Supports multiple image formats (JPG, PNG, WEBP)
• Processes images up to 4096x4096 pixels
• Simple and intuitive UI using Gradio
• Runs on both CPU and GPU

Ensure you have the following installed:

• Python 3.8+
• Pip
• PyTorch (with CUDA if using GPU)

pip install -r requirements.txt

python app.py

This will start a local Gradio server, accessible at http://127.0.0.1:7860/.

1. Upload an image using the input area.
2. Click the "Remove Background" button.
3. Wait for the processing to complete.
4. Download the result with a transparent background.

The BEN_Base model is loaded from Hugging Face:

• Repository: PramaLLC/BEN2
• Model File: BEN2_Base.pth

If the model file is not present, it will be automatically downloaded.

To ensure reproducibility, random seeds are set for:

• Python’s random module
• NumPy
• PyTorch (CPU & CUDA)

1. Load the image and convert it to RGB.
2. Pass the image through the BEN_Base model for inference.
3. Save and return the processed transparent image.

The interface is built with Gradio using a custom modern theme:

• Primary Hue: Indigo
• Secondary Hue: Slate
• Font: Inter, IBM Plex Mono

Contributions are welcome! Feel free to open issues or submit pull requests.

This project is licensed under the MIT License.

• Hugging Face Hub
• Gradio
• PyTorch