CMIB: unsupervised image object categorization in multiple visual contexts
- Submitting institution
-
University of Portsmouth
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 15980429
- Type
- D - Journal article
- DOI
-
10.1109/TII.2019.2939278
- Title of journal
- IEEE Transactions on Industrial Informatics
- Article number
- 0
- First page
- 3974
- Volume
- 16
- Issue
- 6
- ISSN
- 1551-3203
- Open access status
- Compliant
- Month of publication
- September
- 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
-
4
- Research group(s)
-
B - Computational Intelligence
- Citation count
- 0
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This research pushes the boundaries of unsupervised object categorization by developing a novel general-purpose contextual multivariate information bottleneck (CMIB) algorithm. It is the first attempt to utilize the high-level clustering generated by global features as multiple visual contexts, leveraging heterogeneous features and achieving the best categorization performance. Being unsupervised, it is particularly important where image labels are unavailable. This work has led to further funding from Leverhulme Trust (VP1-2020-044, £110,414) for underwater object analysis, reconstruction, and categorization, starting 01/07/2021.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -