Access UniFlow-RS weights
Please provide your full name, institution or organization, and email address. Access is granted automatically after submission. Your responses will be shared with the model authors for access management.
Log in or Sign Up to review the conditions and access this model content.
UniFlow-RS
UniFlow-RS: Unified Generative Prediction for Heterogeneous Remote-Sensing Tasks
Jianlong He, Liang Liao, Jing Xiao, Chaofeng Chen, Gui-Song Xia
Paper · Code · Project page
UniFlow-RS uses a shared generative predictor for building extraction, change detection, semantic segmentation, height estimation, referring segmentation, and rotated object detection.
Download
Sign in to Hugging Face and complete the access form on this model page with your full name, institution or organization, and email address. Access is granted automatically after submission.
Then, from the UniFlow-RS code repository, run:
hf auth login
hf download elaxEgan/UniFlow-RS uniflow_rs.pt --local-dir checkpoints
uniflow_rs.pt contains the shared LoRA adapter, not the full pretrained backbone. It uses the UniFlow-RS checkpoint format and should be loaded with the official code.
Download the base transformer, VAE, and text encoder separately, following the code repository's installation and inference instructions. The release uses FLUX.2-klein-base-4B with a shared adapter and two flow inference steps.
Usage and terms
See the code repository for dataset preparation, evaluation commands, and configuration files. Upstream model components and datasets remain subject to their respective licenses and access requirements.
Citation
@article{he2026uniflowrs,
title={UniFlow-RS: Unified Generative Prediction for Heterogeneous Remote-Sensing Tasks},
author={He, Jianlong and Liao, Liang and Xiao, Jing and Chen, Chaofeng and Xia, Gui-Song},
journal={arXiv preprint arXiv:2611.14281},
year={2026}
}
Model tree for elaxEgan/UniFlow-RS
Base model
black-forest-labs/FLUX.2-klein-base-4B