Image Attribution on CUB-200-2011

Metric: Deletion AUC score (ResNet-101) (higher is better)

LeaderboardDataset

Loading chart...

Results

Submit a result

Sort:

#	Model↕	Deletion AUC score (ResNet-101)▼	Extra Data	Paper	Date↕	Code
1	LIME	0.107	No	"Why Should I Trust You?": Explaining the Predic...	2016-02-16	Code
2	Kernel SHAP	0.1016	No	A Unified Approach to Interpreting Model Predict...	2017-05-22	Code
3	Grad-CAM	0.081	No	Grad-CAM: Visual Explanations from Deep Networks...	2016-10-07	Code
4	Integrated Gradients	0.0728	No	Axiomatic Attribution for Deep Networks	2017-03-04	Code
5	Saliency	0.0682	No	Deep Inside Convolutional Networks: Visualising ...	2013-12-20	Code
6	RISE	0.0665	No	RISE: Randomized Input Sampling for Explanation ...	2018-06-19	Code
7	HSIC-Attribution	0.0647	No	Making Sense of Dependence: Efficient Black-box ...	2022-06-13	Code
8	SMDL-Attribution (ICLR version)	0.0613	No	Less is More: Fewer Interpretable Region via Sub...	2024-02-14	Code

#1LIMESOTA
0.107
Deletion AUC score (ResNet-101)· 2016-02-16
"Why Should I Trust You?": Explaining the Predictions of Any Classifier Code
#2Kernel SHAP
0.1016
Deletion AUC score (ResNet-101)· 2017-05-22
A Unified Approach to Interpreting Model Predictions Code
#3Grad-CAM
0.081
Deletion AUC score (ResNet-101)· 2016-10-07
Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization Code
#4Integrated Gradients
0.0728
Deletion AUC score (ResNet-101)· 2017-03-04
Axiomatic Attribution for Deep Networks Code
#5SaliencySOTA
0.0682
Deletion AUC score (ResNet-101)· 2013-12-20
Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps Code
#6RISE
0.0665
Deletion AUC score (ResNet-101)· 2018-06-19
RISE: Randomized Input Sampling for Explanation of Black-box Models Code
#7HSIC-Attribution
0.0647
Deletion AUC score (ResNet-101)· 2022-06-13
Making Sense of Dependence: Efficient Black-box Explanations Using Dependence Measure Code
#8SMDL-Attribution (ICLR version)
0.0613
Deletion AUC score (ResNet-101)· 2024-02-14
Less is More: Fewer Interpretable Region via Submodular Subset Selection Code