Out-of-Distribution Detection Without Class Labels

Niv Cohen; Ron Abutbul; Yedid Hoshen

RECORD

Out-of-Distribution Detection Without Class Labels

Niv Cohen, Ron Abutbul, Yedid Hoshen

Computer Vision and Pattern Recognition · Submitted 14 Dec 2021

abstract compiled · 2026-09-23 · no language-model review

The authors' main claim We verify the performance of our method by a favorable comparison to the state-of-the-art, and provide extensive analysis and ablations.

how it is stated · asserted 1.00 read the record →

Comment (0) Read on arXiv ↗

Cite ▾

Claim paper

The compiled record

The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Open the compiled record

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 5,419 input tokens.

What the abstract itself states, as probabilities: claims first 0.06 · states limitation 0.09 · human participants 0.03 · controlled comparison 0.69 · randomized assignment 0.02 · states quantitative result 0.04 · compares with prior results 0.96 · mentions code or data availability 0.02.

The abstract, sentence by sentence

arXiv listing abstract v2. Results and the main claim are underlined; show every role.

Out-of-distribution detection seeks to identify novelties, samples that deviate from the norm. The task has been found to be quite challenging, particularly in the case where the normal data distribution consists of multiple semantic classes (e.g., multiple object categories). To overcome this challenge, current approaches require manual labeling of the normal images provided during training. In this work, we tackle multi-class novelty detection without class labels. Our simple but effective solution consists of two stages: we first discover "pseudo-class" labels using unsupervised clustering. Then using these pseudo-class labels, we are able to use standard supervised out-of-distribution detection methods. We verify the performance of our method by a favorable comparison to the state-of-the-art, and provide extensive analysis and ablations.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#	sentence	role	the authors' own work	states the main conclusion	strength
s0	Out-of-distribution detection seeks to identify novelties, samples that deviate from the norm.	background 1.00	0.05	0.03	asserted 0.99
s1	The task has been found to be quite challenging, particularly in the case where the normal data distribution consists of multiple semantic classes (e.g., multiple object categories).	background 0.96	0.19	0.03	asserted 1.00
s2	To overcome this challenge, current approaches require manual labeling of the normal images provided during training.	background 0.53	0.05	0.03	asserted 1.00
s3	In this work, we tackle multi-class novelty detection without class labels.	contribution 0.99	0.84	0.16	asserted 0.95
s4	Our simple but effective solution consists of two stages: we first discover "pseudo-class" labels using unsupervised clustering.	method 0.97	0.91	0.11	asserted 1.00
s5	Then using these pseudo-class labels, we are able to use standard supervised out-of-distribution detection methods.	method 0.99	0.73	0.14	asserted 0.87
s6	We verify the performance of our method by a favorable comparison to the state-of-the-art, and provide extensive analysis and ablations.	result 0.76	0.95	0.38	asserted 1.00

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.89.

What was examined, and what was not

sources read

arXiv listing abstract v2

not examined

full text
figures and tables
references and citing sentences
relations to other papers

levels run

P abstract

not run

S sections
T body sentences
C citations
R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

The paper

Original abstract

Out-of-distribution detection seeks to identify novelties, samples that deviate from the norm. The task has been found to be quite challenging, particularly in the case where the normal data distribution consists of multiple semantic classes (e.g., multiple object categories). To overcome this challenge, current approaches require manual labeling of the normal images provided during training. In this work, we tackle multi-class novelty detection without class labels. Our simple but effective solution consists of two stages: we first discover "pseudo-class" labels using unsupervised clustering. Then using these pseudo-class labels, we are able to use standard supervised out-of-distribution detection methods. We verify the performance of our method by a favorable comparison to the state-of-the-art, and provide extensive analysis and ablations.

Human review

How Pith checks papers → · Full audit of this paper →

How the literature uses it

What this paper does with the arXiv papers it cites

Each citing sentence is taken from the citing paper's LaTeX source and typed by Pith's classifier. A label appears at probability 0.8 or more (0.95 for support), a contest at 0.7 or more; the quoted words are the citing authors'. How each label was checked · JSON

What it cites · 20 sentences without a confident label

Builds on 2 papers

Here, we extend the work of Hendrycks et al. [ref] and Fort et al. [ref] which suggested initializing the classifier $C$ with weights that are pre-trained on a large-scale datasets.

Using Self-Supervised Learning Can Improve Model Robustness and Uncertainty · Feature-Adaptation for Labeled Multi-Class Anomaly Detection · 0.92
Here, we extend the work of Hendrycks et al. [ref] and Fort et al. [ref] which suggested initializing the classifier $C$ with weights that are pre-trained on a large-scale datasets.

Exploring the Limits of Out-of-Distribution Detection · Feature-Adaptation for Labeled Multi-Class Anomaly Detection · 0.92

Uses the method or tool of 3 papers

Recent methods suggested a per-class Mahalanobis approach [ref] deeming a sample normal if it lies within a small Mahalanobis distance to any of the cluster centers.

Exploring the Limits of Out-of-Distribution Detection · Finetuning Pre-trained Features with Pseudo Labels · 0.80
We follow MeanShifted [ref] by using the cosine similarity distance instead of $l_2$ distance.

Mean-Shifted Contrastive Loss for Anomaly Detection · Multi-Class Anomaly Detection With Pre-trained Features · 1.00
Following previous works [ref] , all our main results are reported using $k$NN with $k=1$.

Sub-Image Anomaly Detection with Deep Pyramid Correspondences · Discussion · 0.91

Uses data or values from 1 paper

We evaluate it against a variety of dataset supplied in a similar resolution, namely: CIFAR-100 [ref] , SVHN [ref] , LSUN [ref] .

LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop · Results · 0.88

Compares with 2 papers

We find that our method outperforms state-of-the-art self-supervised [ref] and pre-trained [ref] anomaly detection methods.

Mean-Shifted Contrastive Loss for Anomaly Detection · Introduction · 1.00
For example, in the CIFAR-10 vs. CIFAR-100 multi-class anomaly detection task, our method achieves 96.7% ROCAUC, while a similar method by Fort et al. [ref] achieves $98.4$% ROCAUC using the ground truth labels.

Exploring the Limits of Out-of-Distribution Detection · Discussion · 0.99

Cites as support 1 paper

This phenomenon has also been noticed by previous works [ref] .

Merging Models with Fisher-Weighted Averaging · Discussion · 0.97

Cites as background 4 papers

Different methods suggested modification to this score to account for the dataset statistics (Likelihood Ratio [ref] ) or its complexity (Input Complexity [ref] ).

Likelihood Ratios for Out-of-Distribution Detection · Multi-Class Anomaly Detection Without Pretraining · 0.82
Different methods suggested modification to this score to account for the dataset statistics (Likelihood Ratio [ref] ) or its complexity (Input Complexity [ref] ).

Input complexity and out-of-distribution detection with likelihood-based generative models · Multi-Class Anomaly Detection Without Pretraining · 0.82
Many techniques have been proposed to improve the obtained OOD detection [ref] .

Mitigating Neural Network Overconfidence with Logit Normalization · Feature-Adaptation for Labeled Multi-Class Anomaly Detection · 0.83

Discussion

Know something about this paper? Leave a comment, or highlight any passage above to comment on it.