Enhancing Convolutional Neural Networks for Face Recognition with Occlusion Maps and Batch Triplet Loss
- Submitting institution
-
University of Hertfordshire
- Unit of assessment
- 12 - Engineering
- Output identifier
- 16267233
- Type
- D - Journal article
- DOI
-
10.1016/j.imavis.2018.09.011
- Title of journal
- Image and Vision Computing
- Article number
- -
- First page
- 99
- Volume
- 79
- Issue
- -
- ISSN
- 0262-8856
- Open access status
- Compliant
- Month of publication
- September
- 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
-
2
- Research group(s)
-
-
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- State-of-the-art CNNs rely heavily on specific face regions making them fail to recognise faces with occlusions such as sunglasses or a face mask. This work, from our KTP collaboration with IDscan Biometrics, addresses this by applying occlusion maps strategically, a process that led to the granting of European Patent EP3428843A1. IDscan have gone on to use this approach in their software suite, establishing a dedicated team to convert the KTP outputs to a commercial system. They have gone on to secure contracts with 3 large customers, KBC Bank, Travelex and Credas on the back of this technology.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -