BiRefNet_HR
ZhengPeng7/BiRefNet_HR
A popular open image segmentation model, with 23.5K downloads a month. gigarouter benchmarks and hosts it as an OpenAI-compatible API.
about this model
BiRefNet_HR is a bilateral reference model for high-resolution dichotomous image segmentation (DIS), hosted on gigarouter as a managed API. It is the official implementation of the paper "Bilateral Reference for High-Resolution Dichotomous Image Segmentation" (CAAI AIR 2024). The model is trained on images at 2048×2048 resolution, enabling finer detail capture in segmentation tasks.
Key strengths
- State-of-the-art performance on three tasks: dichotomous image segmentation (DIS), high-resolution salient object detection (HRSOD), and camouflaged object detection (COD).
- Trained at 2048×2048 for high-resolution inference, with FP16 evaluation.
- Particularly effective at segmenting fine structures and thin boundaries in high-resolution images.
Benchmark results (DIS-VD dataset, FP16 mode)
| Method | Resolution | maxFm | wFmeasure | MAE | Smeasure | meanEm | HCE | maxEm | meanFm | adpEm | adpFm | mBA | maxBIoU | meanBIoU |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| BiRefNet_HR-general-epoch_130 | 2048×2048 | .925 | .894 | .026 | .927 | .952 | 811 | .960 | .909 | .944 | .888 | .828 | .837 | .817 |
| BiRefNet_HR-general-epoch_130 | 1024×1024 | .876 | .840 | .041 | .893 | .913 | 1348 | .926 | .860 | .930 | .857 | .765 | .769 | .742 |
| BiRefNet-general-epoch_244 | 2048×2048 | .888 | .858 | .037 | .898 | .934 | 811 | .941 | .878 | .927 | .862 | .802 | .790 | .776 |
| BiRefNet-general-epoch_244 | 1024×1024 | .908 | .877 | .034 | .912 | .943 | 1128 | .953 | .894 | .944 | .881 | .796 | .812 | .789 |
Sample results (DIS-Sample_1 and DIS-Sample_2):
 |
 |
Developed by Peng Zheng, Dehong Gao, Deng-Ping Fan, Li Liu, Jorma Laaksonen, Wanli Ouyang, and Nicu Sebe from Nankai University, Northwestern Polytechnical University, National University of Defense Technology, Aalto University, Shanghai AI Laboratory, and University of Trento. The GitHub repository (https://github.com/ZhengPeng7/BiRefNet) provides codes, documentation, and model zoo.
We're benchmarking and onboarding BiRefNet_HR as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.