Birefnet 512x512

🚀 BiRefNet

BiRefNet is a model designed for image segmentation tasks, including background removal, mask generation, and dichotomous image segmentation. It offers fast and accurate inference at lower resolutions, trained with 512x512 images.

🚀 Quick Start

For more information and usage instructions, please check the main BiRefNet model repo:

• Hugging Face Repo
• GitHub Repo

✨ Features

• Multiple Segmentation Tasks: Suitable for background removal, mask generation, dichotomous image segmentation, camouflaged object detection, and salient object detection.
• Optimized for Lower Resolutions: Trained with 512x512 images for faster and more accurate inference.

📚 Documentation

Performance

All the following tests were conducted in FP16 mode.

Paper and Resources

• Title: Bilateral Reference for High-Resolution Dichotomous Image Segmentation
• Authors:
  • Peng Zheng ^1,4,5,6, Dehong Gao ², Deng-Ping Fan ^1*, Li Liu ³, Jorma Laaksonen ⁴, Wanli Ouyang ⁵, Nicu Sebe ⁶
• Affiliations:
  • ¹ Nankai University, ² Northwestern Polytechnical University, ³ National University of Defense Technology, ⁴ Aalto University, ⁵ Shanghai AI Laboratory, ⁶ University of Trento

                   

Sample Images

DIS-Sample_1	DIS-Sample_2

🔧 Technical Details

This repo is the official implementation of "Bilateral Reference for High-Resolution Dichotomous Image Segmentation" (CAAI AIR 2024).

📄 License

This project is licensed under the MIT license.

Acknowledgement

Many thanks to @freepik for their generous support on GPU resources for training this model!

Citation

@article{zheng2024birefnet,
  title={Bilateral Reference for High-Resolution Dichotomous Image Segmentation},
  author={Zheng, Peng and Gao, Dehong and Fan, Deng-Ping and Liu, Li and Laaksonen, Jorma and Ouyang, Wanli and Sebe, Nicu},
  journal={CAAI Artificial Intelligence Research},
  volume = {3},
  pages = {9150038},
  year={2024}
}