Out Of Distribution Detection On Imagenet 1K 8

평가 지표

AUROC

FPR95

평가 결과

이 벤치마크에서 각 모델의 성능 결과

모델 이름	AUROC	FPR95	Paper Title	Repository
RankFeat (ResNetv2-101)	94.07	29.27	RankFeat: Rank-1 Feature Removal for Out-of-distribution Detection
NPOS	90.44	43.77	Non-Parametric Outlier Synthesis
SHE	-	54.19	Out-of-Distribution Detection based on In-Distribution Data Patterns Memorization with Modern Hopfield Energy
DICE + ReAct (ResNet-50)	93.94	25.45	DICE: Leveraging Sparsification for Out-of-Distribution Detection
ReAct (ResNet-50)	88.16	47.69	ReAct: Out-of-distribution Detection With Rectified Activations
KNN (ResNet-50 SupCon)	88.40	48.91	Out-of-Distribution Detection with Deep Nearest Neighbors
MOS (BiT-S-R101x1)	92.01	40.63	MOS: Towards Scaling Out-of-distribution Detection for Large Semantic Space
MCM (CLIP-L)	94.14	29.00	Delving into Out-of-Distribution Detection with Vision-Language Representations
NNGuide (RegNet)	94.43	21.58	Nearest Neighbor Guidance for Out-of-Distribution Detection
ODIN+UMAP (ResNet-50)	86.92	49.69	Unleashing Mask: Explore the Intrinsic Out-of-Distribution Detection Capability
MCM (CLIP-B)	92.57	37.59	Delving into Out-of-Distribution Detection with Vision-Language Representations
DICE (ResNet-50)	90.83	35.15	DICE: Leveraging Sparsification for Out-of-Distribution Detection
ASH-S (ResNet-50)	94.02	27.98	Extremely Simple Activation Shaping for Out-of-Distribution Detection
MOOD	89.8	-	Rethinking Out-of-distribution (OOD) Detection: Masked Image Modeling is All You Need
SCALE (ResNet50)	95.02	23.27	Scaling for Training Time and Post-hoc Out-of-distribution Detection Enhancement	-
LINe (ResNet50)	95.26	19.48	LINe: Out-of-Distribution Detection by Leveraging Important Neurons
GradNorm (ResNetv2-101)	-	46.48	On the Importance of Gradients for Detecting Distributional Shifts in the Wild
KNN (ResNet-50)	80.10	69.53	Out-of-Distribution Detection with Deep Nearest Neighbors
DOE	76.26	80.94	Out-of-distribution Detection with Implicit Outlier Transformation
DML	-	57.40	Decoupling MaxLogit for Out-of-Distribution Detection	-

0 of 21 row(s) selected.