Skip to content

Latest commit

 

History

7 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 

Repository files navigation

EMMT_SOD

The official repository of "Efficient Multi-Modal Tuning via Visual Prompts for Salient Object Detection." The code and results can be downloaded from this link: https://pan.baidu.com/s/1K3_XnM8iSrmBGwmyVpokjw?pwd=emmt Extraction code: emmt

Results

Quantitative results of different RGB SOD methods

The symbols ↑/↓ indicate that a higher/lower score is better, and the best performances are in bold.

Method Params ↓ PASCAL-S Sm ↑ PASCAL-S Fβω ↑ PASCAL-S MAE ↓ PASCAL-S Em ↑ HKU-IS Sm ↑ HKU-IS Fβω ↑ HKU-IS MAE ↓ HKU-IS Em ↑ DUTS-TE Sm ↑ DUTS-TE Fβω ↑ DUTS-TE MAE ↓ DUTS-TE Em ↑ DUT-OMRON Sm ↑ DUT-OMRON Fβω ↑ DUT-OMRON MAE ↓ DUT-OMRON Em ↑
GateNet 128.6 0.854 0.804 0.071 0.900 0.915 0.880 0.033 0.955 0.885 0.809 0.040 0.928 0.838 0.729 0.055 0.876
MINet 47.6 0.854 0.818 0.066 0.901 0.919 0.897 0.029 0.960 0.884 0.825 0.037 0.927 0.833 0.738 0.056 0.869
VST 44.1 0.871 0.827 0.062 0.918 0.928 0.897 0.029 0.968 0.896 0.828 0.037 0.939 0.850 0.755 0.058 0.888
SGL-KRN 72.4 0.854 0.823 0.070 0.900 0.921 0.904 0.028 0.961 0.893 0.847 0.034 0.939 0.846 0.765 0.049 0.885
ZoomNet 32.4 0.869 0.844 0.057 0.917 0.931 0.918 0.023 0.967 0.900 0.854 0.033 0.936 0.841 0.755 0.053 0.872
MENet -- 0.872 0.847 0.053 0.910 0.927 0.917 0.023 0.960 0.905 0.870 0.028 0.938 0.859 0.785 0.046 0.888
BSNet 53.0 0.842 0.800 0.066 0.897 0.906 0.891 0.031 0.954 0.856 0.797 0.043 0.915 0.815 0.724 0.058 0.863
Ours 4.4 0.881 0.847 0.053 0.920 0.931 0.921 0.024 0.962 0.915 0.881 0.027 0.949 0.862 0.790 0.046 0.893

Quantitative results of different Light Field SOD methods

The symbols ↑/↓ indicate that a higher/lower score is better, and the best performances are in bold.

Method Params ↓ DUTLF-V2 Sm ↑ DUTLF-V2 Fβω ↑ DUTLF-V2 MAE ↓ DUTLF-V2 Em ↑
SA-Net 66.9 0.857 0.792 0.046 0.906
PANet 15.1 0.863 0.785 0.048 0.897
LFBCNet 11.6 0.885 0.821 0.042 0.919
ESCNet 32.8 0.881 0.817 0.042 0.909
CDINet - 0.905 0.861 0.033 0.933
Ours 4.4 0.919 0.884 0.027 0.951

Quantitative results of different RGB-D SOD methods

The symbols ↑/↓ indicate that a higher/lower score is better, and the best performances are in bold.

Method Params ↓ SIP Sm ↑ SIP Fβ ↑ SIP MAE ↓ SIP Em ↑ STERE Sm ↑ STERE Fβ ↑ STERE MAE ↓ STERE Em ↑ NJU2K Sm ↑ NJU2K Fβ ↑ NJU2K MAE ↓ NJU2K Em ↑
CMINet 196.2 0.899 0.872 0.040 0.937 0.918 0.886 0.032 0.948 0.929 0.910 0.029 0.953
UCNet-CVAE 31.3 0.882 0.850 0.045 0.927 0.906 0.878 0.036 0.945 0.904 0.886 0.038 0.943
DCF 12.0 0.876 0.838 0.052 0.915 0.902 0.867 0.039 0.902 0.912 0.886 0.036 0.944
SSLSOD 74.2 0.868 0.830 0.058 0.903 0.885 0.845 0.047 0.885 0.901 0.872 0.042 0.932
CCFENet 28.7 0.882 0.857 0.047 0.920 0.906 0.881 0.035 0.906 0.917 0.897 0.032 0.949
DIGR-Net 166.7 0.885 0.841 0.052 0.913 0.916 0.870 0.037 0.916 0.933 0.904 0.028 0.953
CAVER 55.8 0.893 0.864 0.042 0.933 0.913 0.882 0.033 0.949 0.921 0.901 0.031 0.953
LSNet 4.6 0.886 0.847 0.050 0.920 0.871 0.816 0.055 0.908 0.911 0.878 0.039 0.939
Ours 4.4 0.908 0.884 0.036 0.908 0.924 0.894 0.030 0.951 0.929 0.907 0.029 0.953

Quantitative results of different RGB-T SOD methods

The symbols ↑/↓ indicate that a higher/lower score is better, and the best performances are in bold.

Method Params ↓ VT821 Sm ↑ VT821 Fβω ↑ VT821 MAE ↓ VT821 Em ↑ VT1000 Sm ↑ VT1000 Fβω ↑ VT1000 MAE ↓ VT1000 Em ↑ VT5000 Sm ↑ VT5000 Fβω ↑ VT5000 MAE ↓ VT5000 Em ↑
ECFFNet - 0.877 0.799 0.035 0.907 0.924 0.883 0.022 0.910 0.875 0.800 0.038 0.910
CGFNet 69.9 0.880 0.829 0.038 0.918 0.923 0.900 0.023 0.955 0.883 0.831 0.035 0.924
APNet 30.4 0.868 0.791 0.034 0.898 0.922 0.882 0.022 0.949 0.876 0.805 0.035 0.913
CCFENet 28.7 0.900 0.852 0.027 0.932 0.934 0.910 0.018 0.964 0.896 0.849 0.030 0.935
TNet 87.0 0.899 0.841 0.030 0.928 0.929 0.895 0.021 0.957 0.895 0.840 0.033 0.931
LSNet 4.6 0.879 0.808 0.033 0.911 0.926 0.886 0.023 0.954 0.877 0.804 0.037 0.916
CAVER 55.8 0.891 0.835 0.033 0.926 0.936 0.909 0.017 0.965 0.892 0.835 0.032 0.930
TMMANet 141 0.880 0.804 0.032 0.908 0.939 0.908 0.018 0.946 0.898 0.836 0.030 0.929
Ours 4.4 0.901 0.843 0.034 0.925 0.941 0.916 0.017 0.967 0.901 0.853 0.028 0.936

BibTeX

@article{zhang2025parameter,
  title={Efficient Multi-Modal Tuning via Visual Prompts for Salient Object Detection},
  author={Zhang, Zixuan and Shi, Fan and Jia, Chen and Wang, Mianzhao and Louis, Assale Adje and Cheng, Xu},
  journal={ACM Transactions on Multimedia Computing, Communications, and Applications},
  year={2025},
}

About

The official repository of "Efficient Multi-Modal Tuning via Visual Prompts for Salient Object Detection."

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors