Deep Bayesian Self-Training
- Submitting institution
-
University of Aberdeen
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 158824358
- Type
- D - Journal article
- DOI
-
10.1007/s00521-019-04332-4
- Title of journal
- Neural Computing and Applications
- Article number
- -
- First page
- 4275
- Volume
- 32
- Issue
- -
- ISSN
- 0941-0643
- Open access status
- Compliant
- Month of publication
- July
- Year of publication
- 2019
- URL
-
-
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
5
- Research group(s)
-
-
- Citation count
- 1
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- Adopting Machine Learning (ML) at scale requires huge amounts of labelled data, which is arguably a very expensive process. However, the method proposed in this paper enables the self-training of deep learning systems through a novel self-annotation process. This paper is significant because it is the first paper to show how Bayesian deep learning can be used to reliably self-annotate large amounts of data based on a novel criterion. The proposed method was evaluated in a large bespoke food packaging image dataset and used in a food packaging supply chain for quality assurance via a collaboration with Olympus Automation Ltd.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -
