Bayesian Joint Modelling for Object Localisation in Weakly Labelled Images
- Submitting institution
-
Queen Mary University of London
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 2557
- Type
- D - Journal article
- DOI
-
10.1109/TPAMI.2015.2392769
- Title of journal
- IEEE Transactions on Pattern Analysis and Machine Intelligence
- Article number
- 10
- First page
- 1959
- Volume
- 37
- Issue
- 10
- ISSN
- 0162-8828
- Open access status
- Out of scope for open access requirements
- Month of publication
- October
- Year of publication
- 2015
- 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
-
2
- Research group(s)
-
-
- Citation count
- 13
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This paper solves the popular problem of object detection with cluttered background but under a different setting: only the presence of the objects is known without their precise location. Solving this problem would mean that the abundant images on social media sites with loosely related user tags can be used to train object detection model. To solve this weakly supervised learning problem, a generative Bayesian joint topic model is proposed. The concept has inspired the multi-instance learning based object localisation work at INRIA (Cinbis et al, IEEE Trans. PAMI?16, 10.1109/TPAMI.2016.2535231).
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -