Out Of Distribution Detection
Out-of-Distribution Detection refers to identifying anomalous samples that do not belong to the distribution of training data in computer vision tasks. This task aims to enhance the robustness and generalization ability of models, effectively avoiding misjudgments on unknown data by detecting and filtering these anomalies, thus improving the safety and reliability of the system. In practical applications, this technology is crucial for boosting system performance in fields such as autonomous driving and medical image analysis.
20 Newsgroups
2-Layered GRU
ADE-OoD
RbA
CIFAR-10
Wide ResNet 40x2
CIFAR-10 vs CIFAR-10.1
ERD (ResNet18)
CIFAR-10 vs CIFAR-100
Wide 40-2 + OECC
CIFAR-10 vs Gaussian
CIFAR-10 vs ImageNet (C)
CIFAR-10 vs ImageNet (R)
CIFAR-10 vs iSUN
CIFAR-10 vs LSUN (C)
CIFAR-10 vs LSUN (R)
CIFAR-10 vs SVHN
CIFAR-10 vs Uniform
CIFAR-100
Wide ResNet 40x2
CIFAR-100 vs CIFAR-10
WRN 40-2 + OECC
CIFAR-100 vs Gaussian
CIFAR-100 vs ImageNet (C)
CIFAR-100 vs ImageNet (R)
DenseNet-BC-100
CIFAR-100 vs iSUN
DenseNet-BC-100
CIFAR-100 vs LSUN (C)
CIFAR-100 vs LSUN (R)
DenseNet-BC-100
CIFAR-100 vs SVHN
OECC + MD
CIFAR-100 vs Uniform
cifar10
cifar100
Wide Resnet 40x2
Far-OOD
ISH (ResNet50)
Fashion-MNIST
PAE
ImageNet-1k vs Curated OODs (avg.)
NNGuide (RegNet)
ImageNet-1K vs ImageNet-C
ImageNet-1K vs ImageNet-O
NNGuide-ViM (ViT-B/16)
ImageNet-1k vs iNaturalist
NNGuide (RegNet)
ImageNet-1k vs NINCO
Forte
ImageNet-1k vs Places
BATS (ResNet-50)
ImageNet-1K vs SSB-hard
ImageNet-1k vs SUN
LINe (ResNet50)
ImageNet-1k vs Textures
ViM (BiT-S-R101×1)
ImageNet dogs vs ImageNet non-dogs
ResNet34 + FSSD
ImageNet-1k vs OpenImage-O
NNGuide (RegNet)
MS-1M vs. IJB-C
ResNeXt50 + FSSD
Near-OOD
SST
STL-10
Mixup (Gaussian)
SVHN vs CIFAR-10
SVHN vs CIFAR-100
SVHN vs Gaussian
SVHN vs ImageNet (C)
SVHN vs ImageNet (R)
SVHN vs iSUN
SVHN vs LSUN (C)
SVHN vs LSUN (R)
SVHN vs Uniform
Wide ResNet 40x2