Out Of Distribution Detection On Imagenet 1K 10

평가 지표

AUROC

FPR95

평가 결과

이 벤치마크에서 각 모델의 성능 결과

모델 이름	AUROC	FPR95	Paper Title	Repository
DICE + ReAct (ResNet-50)	92.74	28.07	DICE: Leveraging Sparsification for Out-of-Distribution Detection
GradNorm (ResNetv2-101)	-	61.42	On the Importance of Gradients for Detecting Distributional Shifts in the Wild
LINe (ResNet-50)	94.44	22.54	LINe: Out-of-Distribution Detection by Leveraging Important Neurons
Watermarking (WRN-40-2 w/ MSP)	84.00	61.2	Watermarking for Out-of-distribution Detection
Watermarking (WRN-40-2 w/ Energy)	80.80	67.75	Watermarking for Out-of-distribution Detection
MOS (BiT-S-R101x1)	81.23	60.43	MOS: Towards Scaling Out-of-distribution Detection for Large Semantic Space
ViM (ViT-B/16)	95.34	20.31	ViM: Out-Of-Distribution with Virtual-logit Matching
ODIN+UMAP (ResNet-50)	88.35	42.02	Unleashing Mask: Explore the Intrinsic Out-of-Distribution Detection Capability
DOE	88.9	34.67	Out-of-distribution Detection with Implicit Outlier Transformation
GEN	-	77.93	GEN: Pushing the Limits of Softmax-Based Out-of-Distribution Detection
ViM (BiT-S-R101×1)	-	4.69	ViM: Out-Of-Distribution with Virtual-logit Matching
NAC-UE (ResNet-50)	97.9	-	Neuron Activation Coverage: Rethinking Out-of-distribution Detection and Generalization
BATS (ResNet-50)	92.27	38.9	Boosting Out-of-distribution Detection with Typical Features	-
SSD	85.4	57.2	SSD: A Unified Framework for Self-Supervised Outlier Detection
KNN (ResNet-50 SupCon)	94.45	16.97	Out-of-Distribution Detection with Deep Nearest Neighbors
NNGuide (RegNet)	95.82	17.00	Nearest Neighbor Guidance for Out-of-Distribution Detection
MOOD	91.3	-	Rethinking Out-of-distribution (OOD) Detection: Masked Image Modeling is All You Need
RP+GradNorm	-	79.80	Detecting Out-of-distribution Data through In-distribution Class Prior
KNN (ResNet-50)	97.18	11.56	Out-of-Distribution Detection with Deep Nearest Neighbors
SHE	-	45.09	Out-of-Distribution Detection based on In-Distribution Data Patterns Memorization with Modern Hopfield Energy

0 of 33 row(s) selected.