Is it all a cluster game? -- Exploring Out-of-Distribution Detection based on Clustering in the Embedding Space

Sinhamahapatra, Poulami; Koner, Rajat; Roscher, Karsten; Günnemann, Stephan

doi:10.48550/arxiv.2203.08549

Cited by 1 publication

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Since images from a same class will be similar to each other's augmentations in Self-SSL, they will cluster together in the feature space to create one (or more) mode for the class (the alignment principle). Based on this, recently published works have discovered that classifiers learned by self-supervised image clustering methods provide a strong baseline for out-of-distribution detection on unlabeled multi-class datasets [49], [50]. Inspired by these, we propose a novel contrastive clustering method to identify out-of-distribution data.…”

Section: Detection Of Out-of-distribution Data Using Sc-based Clusteringmentioning

confidence: 97%

Rethinking Pseudo-Labeling for Semi-Supervised Facial Expression Recognition With Contrastive Self-Supervised Learning

Fang

Han

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Self-supervised learning for semi-supervised facial expression recognition aims to avoid the need to collect expensive labeled facial expression data. Existing methods demonstrate an impressive performance boost, but they artificially assume that small labeled facial expression data and large unlabeled facial expression data are from the same class distribution. In a more realistic scenario, when utilizing facial expression data from a large face recognition database as unlabeled data, there will be a mismatch distribution between the two sets of data. This often results in severe performance degradation due to incorrect propagation of unlabeled data from unrelated sources. In this paper, we address the class distribution mismatch problem in deep semi-supervised learning-based facial expression recognition. Specifically, we propose a silhouette coefficient-based contrast clustering algorithm, which determines the degree of separation between clusters by examining the intra-cluster and inter-cluster distances to accurately detect out-of-distribution data. In addition, we propose a pseudo-labeling rethinking strategy that matches the soft pseudo-labels estimated from a fine-tuned network to the contrast clustering to produce reliable pseudolabels. Experiments on three in-the-wild datasets, RAF-DB, FERPlus and AffectNet, demonstrate the effectiveness of our method, and our approach performs well compared to state-of-the-art methods.INDEX TERMS Facial expression recognition, semi-supervised learning, contrastive self-supervised learning, out-of-distribution data.

show abstract

Section: Detection Of Out-of-distribution Data Using Sc-based Clusteringmentioning

confidence: 97%

Rethinking Pseudo-Labeling for Semi-Supervised Facial Expression Recognition With Contrastive Self-Supervised Learning

Fang

Han

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

Is it all a cluster game? -- Exploring Out-of-Distribution Detection based on Clustering in the Embedding Space

Cited by 1 publication

References 0 publications

Rethinking Pseudo-Labeling for Semi-Supervised Facial Expression Recognition With Contrastive Self-Supervised Learning

Rethinking Pseudo-Labeling for Semi-Supervised Facial Expression Recognition With Contrastive Self-Supervised Learning

Contact Info

Product

Resources

About