Pith. sign in

REVIEW

Multi-view Deep One-class Classification: A Systematic Exploration

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2104.13000 v1 pith:N4KLUD4P submitted 2021-04-27 cs.CV cs.LGcs.MM

Multi-view Deep One-class Classification: A Systematic Exploration

classification cs.CV cs.LGcs.MM
keywords multi-viewdeepbenchmarkdatadatasetsavailablebeenclass
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

One-class classification (OCC), which models one single positive class and distinguishes it from the negative class, has been a long-standing topic with pivotal application to realms like anomaly detection. As modern society often deals with massive high-dimensional complex data spawned by multiple sources, it is natural to consider OCC from the perspective of multi-view deep learning. However, it has not been discussed by the literature and remains an unexplored topic. Motivated by this blank, this paper makes four-fold contributions: First, to our best knowledge, this is the first work that formally identifies and formulates the multi-view deep OCC problem. Second, we take recent advances in relevant areas into account and systematically devise eleven different baseline solutions for multi-view deep OCC, which lays the foundation for research on multi-view deep OCC. Third, to remedy the problem that limited benchmark datasets are available for multi-view deep OCC, we extensively collect existing public data and process them into more than 30 new multi-view benchmark datasets via multiple means, so as to provide a publicly available evaluation platform for multi-view deep OCC. Finally, by comprehensively evaluating the devised solutions on benchmark datasets, we conduct a thorough analysis on the effectiveness of the designed baselines, and hopefully provide other researchers with beneficial guidance and insight to multi-view deep OCC. Our data and codes are opened at https://github.com/liujiyuan13/MvDOCC-datasets and https://github.com/liujiyuan13/MvDOCC-code respectively to facilitate future research.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.