Out-of-Distribution Detection

326 papers with code • 50 benchmarks • 22 datasets

Detect out-of-distribution or anomalous examples.

Benchmarks

Add a Result

These leaderboards are used to track progress in Out-of-Distribution Detection

Dataset	Best Model	Compare
ImageNet-1k vs Textures	ViM (BiT-S-R101×1)	See all
ImageNet-1k vs iNaturalist	NNGuide (RegNet)	See all
ImageNet-1k vs Places	LINe (ResNet50)	See all
ImageNet-1k vs SUN	LINe (ResNet50)	See all
ImageNet-1k vs Curated OODs (avg.)	NNGuide (RegNet)	See all
CIFAR-100 vs CIFAR-10	DHM	See all
CIFAR-10 vs CIFAR-100	DHM	See all
CIFAR-10	DHM	See all
STL-10	Mixup (Imagenet)	See all
CIFAR-100 vs SVHN	OECC + MD	See all
ImageNet-1k vs OpenImage-O	NNGuide (RegNet)	See all
CIFAR-100	Wide ResNet 40x2	See all
MS-1M vs. IJB-C	ResNeXt50 + FSSD	See all
ImageNet dogs vs ImageNet non-dogs	ResNet34 + FSSD	See all
ImageNet-1K vs ImageNet-O	NNGuide-ViM (ViT-B/16)	See all
ImageNet-1k vs NINCO	ViT-B-384 Mahalanobis (pre-trained on IN-21k)	See all
20 Newsgroups	2-Layered GRU	See all
Fashion-MNIST	PAE	See all
CIFAR-100 vs iSUN	DenseNet-BC-100	See all
CIFAR-100 vs LSUN (C)	ResNet-34	See all
CIFAR-100 vs LSUN (R)	DenseNet-BC-100	See all
CIFAR-100 vs ImageNet (R)	DenseNet-BC-100	See all
CIFAR-100 vs ImageNet (C)	DenseNet-BC-100	See all
CIFAR-100 vs Uniform	DenseNet-BC-100	See all
CIFAR-100 vs Gaussian	DenseNet-BC-100	See all
CIFAR-10 vs SVHN	DenseNet-BC-100	See all
CIFAR-10 vs LSUN (C)	DenseNet-BC-100	See all
SVHN vs LSUN (C)	DenseNet-BC-100	See all
CIFAR-10 vs iSUN	DenseNet-BC-100	See all
CIFAR-10 vs ImageNet (R)	DenseNet-BC-100	See all
CIFAR-10 vs ImageNet (C)	DenseNet-BC-100	See all
CIFAR-10 vs Uniform	DenseNet-BC-100	See all
CIFAR-10 vs Gaussian	DenseNet-BC-100	See all
SVHN vs iSUN	DenseNet-BC-100	See all
SVHN vs LSUN (R)	DenseNet-BC-100	See all
SVHN vs ImageNet (R)	DenseNet-BC-100	See all
SVHN vs ImageNet (C)	DenseNet-BC-100	See all
SVHN vs Uniform	DenseNet-BC-100	See all
SVHN vs Gaussian	DenseNet-BC-100	See all
SVHN vs CIFAR-10	DenseNet-BC-100	See all
SVHN vs CIFAR-100	DenseNet-BC-100	See all
CIFAR-10 vs LSUN (R)	ResNet-34	See all
Near-OOD	ISH (ResNet50)	See all
Far-OOD	ISH (ResNet50)	See all
TREC-NEWS	2-Layered GRU	See all
SST	2-Layered GRU	See all
CIFAR-10 vs CIFAR-10.1	ERD (ResNet18)	See all
CIFAR10	Wide ResNet 40x2	See all
cifar10	Wideresnet 40	See all
cifar100	Wide Resnet 40x2	See all

Show all 50 benchmarks

Collapse benchmarks

Libraries

Use these libraries to find Out-of-Distribution Detection models and implementations

jingkang50/openood

6 papers

749

kingjamessong/rankfeat

5 papers

lancopku/avg-avg

5 papers

dlmacedo/entropic-out-of-distributi…

4 papers

See all 8 libraries.

Datasets

Most implemented papers

Most implemented Social Latest No code

Hierarchical VAEs Know What They Don't Know

vlievin/biva-pytorch • • 16 Feb 2021

Deep generative models have been demonstrated as state-of-the-art density estimators.

Paper
Code

MOS: Towards Scaling Out-of-distribution Detection for Large Semantic Space

deeplearning-wisc/large_scale_ood • • CVPR 2021

Detecting out-of-distribution (OOD) inputs is a central challenge for safely deploying machine learning models in the real world.

Paper
Code

Open-set Label Noise Can Improve Robustness Against Inherent Label Noise

hongxin001/ODNL • • NeurIPS 2021

Learning with noisy labels is a practically challenging problem in weakly supervised learning.

Paper
Code

A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification

aangelopoulos/conformal-prediction • • 15 Jul 2021

Conformal prediction is a user-friendly paradigm for creating statistically rigorous uncertainty sets/intervals for the predictions of such models.

Paper
Code

Using Self-Supervised Learning Can Improve Model Robustness and Uncertainty

hendrycks/ss-ood • • NeurIPS 2019

Self-supervision provides effective representations for downstream tasks without requiring labels.

Paper
Code

Natural Adversarial Examples

hendrycks/natural-adv-examples • • CVPR 2021

We also curate an adversarial out-of-distribution detection dataset called ImageNet-O, which is the first out-of-distribution detection dataset created for ImageNet models.

Paper
Code

Scaling Out-of-Distribution Detection for Real-World Settings

hendrycks/anomaly-seg • • 25 Nov 2019

We conduct extensive experiments in these more realistic settings for out-of-distribution detection and find that a surprisingly simple detector based on the maximum logit outperforms prior methods in all the large-scale multi-class, multi-label, and segmentation tasks, establishing a simple new baseline for future work.

Paper
Code

Training Normalizing Flows with the Information Bottleneck for Competitive Generative Classification

VLL-HD/FrEIA • • NeurIPS 2020

In this work, firstly, we develop the theory and methodology of IB-INNs, a class of conditional normalizing flows where INNs are trained using the IB objective: Introducing a small amount of {\em controlled} information loss allows for an asymptotically exact formulation of the IB, while keeping the INN's generative capabilities intact.

Paper
Code