Yanguang Sun, Jiawei Lian, Jian Yang, Lei Luo

Our work has been accepted for ICCV 2025. The code has already been open sourced.

If you are interested in our work, please do not hesitate to contact us at Sunyg@njust.edu.cn via email.

Segmentation results

We provide the predicted results genereted by our Controllable-LPMoE model across six binary object segmentation tasks, including ''Camouflaged Object Detection (COD)'', ''Salient Object Detection (SOD)'', ''Polyp Segmentation (PS)'', ''Skin Lesion Segmentation (SLS)'', ''Shadow Detection (SD)'', and ''Glass Detection (GD)''.

LPMoE_U_B_ICCV25_COD [(https://pan.baidu.com/s/1KABwnsRhw75Wecya1RQ6Fw?pwd=bicb), PIN:bicb]

LPMoE_U_B_ICCV25_SOD [(https://pan.baidu.com/s/18Hf5KTqyZliLgv30qGEJvw?pwd=ysst), PIN:ysst]

LPMoE_U_B_ICCV25_PS [(https://pan.baidu.com/s/1gwPD7ti9OnpGuyOIgB_oeQ), PIN:2nff]

LPMoE_U_B_ICCV25_SLS [(https://pan.baidu.com/s/1LZuOXTFBHmo6ka3RhzfRTg), PIN:zpbh]

LPMoE_U_B_ICCV25_SD [(https://pan.baidu.com/s/1kpTEFNSSYqBCW7bRwEpApQ), PIN:csim]

LPMoE_U_B_ICCV25_GD [(https://pan.baidu.com/s/1CcD3AAKMSTiYo3HHWoFtrg), PIN:yxqc]

Training

To train Controllable_LPMoE on COD on a single node with 4 gpus run:

bash dist_train.sh configs/COS/Controllable_LPMoE_COD_Beit.py 4

Testing

The visual segmentation results can be obtained through image_demo.py

Citation

If you use Controllable-LPMoE in your research or wish to refer to the baseline results, please use the following BibTeX entry.

@inproceedings{Controllable-LPMoE,
  title={Controllable-LPMoE: Adapting to Challenging Object Segmentation via Dynamic Local Priors from Mixture-of-Experts},
  author={Sun, Yanguang and Lian, Jiawei and Yang, Jian and Luo, Lei},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={22327--22337},
  year={2025}
}