Dr Sadi Vural
Dr Sadi Vural is the founder and Chief Executive Officer of Ayonix, the Tokyo face recognition company he started in 2007. He holds a doctorate from Osaka University, where his 2011 dissertation addressed face recognition for outdoor surveillance, and his published work covers illumination-invariant recognition and object detection in uncontrolled conditions.
What is Dr Sadi Vural's role at Ayonix?
He founded Ayonix in Tokyo in 2007 and is its Chief Executive Officer. He also writes and signs the technical articles this site publishes on face recognition evaluation, accuracy and deployment, which is why the articles carry his name rather than a company byline: a reader who wants to weigh a claim about biometric accuracy should be able to see who is making it and what they have published.
The company's name comes from that research. The 2011 paper on illumination normalisation introduced a filter bank the authors called Ayofa-Filters, and the product line that grew out of that work took the name Ayonix.
What is his academic background?
Dr Vural completed his doctorate at Osaka University in 2011. The dissertation, "Face recognition by using hybrid-holistic methods for outdoor surveillance systems", is held in the Osaka University Knowledge Archive, the university's own institutional repository, and is linked below rather than summarised here.
His research at Osaka University was carried out with Yasushi Mae, Kenichi Ohara and Tatsuo Arai, and the resulting papers appeared in Pattern Recognition Letters, Machine Vision and Applications and the Journal of Robotics and Mechatronics. Earlier work at Ritsumeikan University addressed digital watermarking for cinema content.
- Doctoral dissertation, Osaka University (2011)Abstract, in the Osaka University Knowledge Archive.
What has he published on face recognition?
The work below is peer-reviewed or archived, and each entry links to its publisher record. The through-line is the problem that still decides whether a deployment works: recognising a face when the lighting, the angle and the distance are not chosen by the system.
- Face relighting using discriminative 2D spherical spaces for face recognition (2013)Machine Vision and Applications.
- Spherical Spaces for Illumination Invariant Face Relighting (2013)Journal of Robotics and Mechatronics.
- Multi-view fast object detection by using extended haar filters in uncontrolled environments (2011)Pattern Recognition Letters.
- Illumination Normalization for Outdoor Face Recognition by Using Ayofa-Filters (2011)Journal of Pattern Recognition Research.
- Robust Digital Cinema Watermarking (2008)Ritsumeikan University.
This list covers published research, not product performance. Nothing in it describes how a current Ayonix system behaves at a particular site, and it should not be read as a benchmark.
What does he write about on this site?
The technical articles here answer the questions that decide a biometric purchase and are usually answered badly: what an independent evaluation does and does not tell you, why a single accuracy percentage is not portable between deployments, and where recognition should run when the data cannot leave the site.
One point recurs across those articles and is worth stating here too: NIST evaluates face recognition algorithms and publishes comparative results, but it does not certify, approve or endorse vendors. No such certification exists, and any company describing itself as NIST certified is describing something that is not real.
- Ayonix in NIST face recognition evaluationsWhat the evaluation measures, and what participation does not mean.
- Why this site publishes no accuracy percentageAnd what to ask a vendor whose answer to every question is one.
- Best face recognition software and companiesA comparison Ayonix publishes and appears in, with the disclosure that requires.
Editorial policy and conflicts of interest
Dr Vural is the founder and CEO of Ayonix, so anything he writes here about face recognition is written by an interested party. That is disclosed on each article as well as here. The site's standing rules are the practical form of that disclosure: no accuracy figure is published for any vendor including Ayonix, every statistic cites a government or academic source, and no claim about a customer appears without that customer's agreement.
Corrections are welcome and are made in public: the page's "last updated" date changes when the text does. Write to infojp@ayonix.com with the page and the correction.
Frequently asked questions
- Who is Dr Sadi Vural?
- He is the founder and Chief Executive Officer of Ayonix, a face recognition and enterprise AI company he started in Tokyo in 2007. He holds a doctorate from Osaka University, where his 2011 dissertation addressed face recognition for outdoor surveillance systems.
- Where did Dr Sadi Vural earn his doctorate?
- At Osaka University. The dissertation, "Face recognition by using hybrid-holistic methods for outdoor surveillance systems", is held in the Osaka University Knowledge Archive, the university's institutional repository, and is linked from this page.
- What has he published?
- Peer-reviewed work on illumination-invariant face recognition and object detection, in Pattern Recognition Letters, Machine Vision and Applications and the Journal of Robotics and Mechatronics, alongside earlier work on digital cinema watermarking at Ritsumeikan University. Each paper is linked to its publisher record on this page.
- Is Ayonix named after his research?
- Yes. His 2011 paper on illumination normalisation for outdoor face recognition introduced a filter bank called Ayofa-Filters, and the company that grew out of that line of work took the name Ayonix.
- Does he develop face recognition in Japan?
- Yes. He develops face recognition technology in Tokyo, Japan, and the work is not limited to recognition itself: it covers the pipeline from face detection through feature extraction, and extends to face tracking, facial expression analysis, gender and age analysis, race analysis and 3D face frontalisation.