Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Early High-Frequency Injection for Geometry-Sensitive OOD Detection

About

Post-hoc OOD detectors score logits or features after training, so their success depends on the geometry already encoded in the representation. We revisit this assumption through a band-wise MMD^2 analysis across CE, SimCLR, SupCon, and the OOD-oriented representation method PALM. In our diagnostic, low-frequency input bands induce weaker ID/OOD feature discrepancy, whereas higher-frequency bands tend to provide stronger separability. This observation motivates EIHF, an input-side intervention that exposes high-frequency evidence before the first convolution without changing the training objective. EIHF is strongest for geometry-sensitive OOD detection: under matched training and scoring settings, it reshapes class-conditional feature geometry and reduces ID/OOD Mahalanobis score overlap. Experiments on CIFAR-100 and ImageNet-100 show gains on CIFAR-100 and the best average FPR95 with second-best average AUROC on ImageNet-100, while also revealing a limitation on the scene-centric Places shift. Code is available at https://anonymous.4open.science/r/EIHF.

Chuanjie Cheng, Ningkang Peng, Chenxi Liu, Yifan He, Peirong Ma, Yanhui Gu• 2026

Related benchmarks

TaskDatasetResultRank
OOD DetectionCIFAR-100 vs ISUN (test)
FPR @ 0.05 FNR14.14
41
OOD DetectionCIFAR-100 LSUN IND R OOD
AUROC97.59
38
OOD DetectionCIFAR-10
FPR95 (SVHN)1.36
36
OOD DetectionImageNet-100 (ID) vs TEXTURES (OOD)
AUROC99.58
33
Out-of-Distribution DetectionCIFAR100 (ID) vs SVHN (OOD)
AUROC0.9908
33
Out-of-distribution (OOD) detectionCIFAR100 In-Distribution Place365 Out-of-Distribution (test)
AUROC83.46
29
OOD DetectionCIFAR-100 OOD (Average of SVHN, Places365, LSUN, iSUN, Textures)
FPR@9525.73
14
OOD DetectionCIFAR-100 (In-distribution) vs Textures (OOD)
FPR9538.78
14
OOD DetectionImageNet100 (In-distribution) vs SUN (OOD) (test)
FPR@9523.83
14
OOD DetectionImageNet-100
Avg FPR9552.03
14
Showing 10 of 12 rows

Other info

Follow for update