Mixed Supervised Object Detection with Robust Objectness Transfer
- Submitting institution
-
University of Dundee
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 35098008
- Type
- D - Journal article
- DOI
-
10.1109/TPAMI.2018.2810288
- Title of journal
- IEEE Transactions on Pattern Analysis and Machine Intelligence
- Article number
- -
- First page
- 639
- Volume
- 41
- Issue
- 3
- ISSN
- 0162-8828
- Open access status
- Exception within 3 months of publication
- Month of publication
- February
- Year of publication
- 2018
- 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
-
3
- Research group(s)
-
-
- Citation count
- 2
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- A robust objectness transfer approach is proposed to leverage object categories with labelled data so as to learn to detect objects from classes with weakly labelled data; it can learn domain-invariant objectness and reject distractors. State-of-the-art performance is achieved on the widely used ILSVRC2013 and PASCAL VOC datasets. This work inspired the development of several systems, including attention-based webly-supervised object detection (Wu et al., CVPR 2020) involving NVIDIA, one-shot segmentation (Zhao et al., ECCV 2020) involving Adobe Research, and a combination with reinforcement learning (Zhang et al., IEEE Trans. NNLS, 2020).
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -