README.md

ISNet Background Removal Model

This repository contains the ISNet (Image Segmentation Network) model for high-quality background removal.

Model Information

• Architecture: ISNetDIS (ISNet for Image Segmentation)
• Input: RGB images (3 channels)
• Output: Segmentation masks (1 channel)
• Model Size: ~168MB
• Framework: PyTorch

Usage

from models import ISNetDIS
import torch

# Load model
model = ISNetDIS()
state_dict = torch.load("supplyswap_isnet.pth", map_location="cpu")
model.load_state_dict(state_dict)
model.eval()

# Use for inference
# See the Spaces demo for complete usage examples

Files

• supplyswap_isnet.pth: Pre-trained model weights
• models/: Model architecture implementation
  • isnet.py: ISNet model implementation
  • __init__.py: Model imports

Citation

If you use this model, please cite the original ISNet paper.

License

Please refer to the original model's license for usage terms.