In short
Every claim this site makes, with the source that supports it and what it does not establish. Claims are registered before they may be published, and the build fails if a page states one it is not approved for. Rejected claims are listed with the reason.
26
approved claims, each with a source and a named reviewer
3
rejected after review, with the reason recorded
20
phrases the build refuses to publish, in any casing
Every high-risk factual claim is declared in src/data/verified-claims.ts before it
may appear on a page. A declaration carries the claim text, its source and source type, a verification
date, a named reviewer, the exact routes it is approved for, and — importantly — what it does
not establish.
The components that render evidence throw at build time if a page uses a claim it is not approved for, or one whose status is anything other than approved. A separate quality gate scans the built HTML for banned phrases and for high-risk language on pages with no corresponding approval. A page cannot ship with an unsupported claim; the build fails first.
High-risk categories are: accuracy, speed, scale, countries, customers, government deployments, awards, certifications, standards compliance, partner status, hardware specifications and platform support.
Approved
Grouped by category. Every entry shows its source, the date it was verified, who verified it, and the limits of what it establishes.
Ayonix algorithms have been submitted to NIST face recognition evaluations
What this does not establish
NIST publishes comparative evaluations and does not endorse, certify, approve or accredit vendors. Participation means an algorithm was measured on data the submitter did not choose. It is not a statement about performance at any particular site.
One verification algorithm, ayonix_000, submitted to the NIST 1:1 verification track on 22 June 2017
What this does not establish
No error rate or ranking from the report is quoted. A 1:1 result does not predict 1:N behaviour.
Three identification algorithms, ayonix_0, ayonix_1 and ayonix_2, with traces in the NIST 1:N identification track from 2018
Ayonix did not participate in the standalone NIST FRVT demographic effects study
What this does not establish
Demographic figures inside the ayonix_000 report card are part of the standard NIST template and are not a demographic-study result.
Ayonix algorithm ayonix_001 is listed in the NIST FATE Age Estimation & Verification evaluation, submitted 23 July 2026
What this does not establish
Participation, not a result: NIST evaluates algorithms and does not endorse, certify or approve vendors. No error figure from the report card is quoted on this site; the report card is linked as the source. Age estimation is a distinct task from the 1:1 and 1:N recognition tracks, and its figures do not transfer to them.
Founded in Tokyo in 2007
Ayonix Co., Ltd.
What this does not establish
A first-party statement, not a value read from a corporate registry. Registry documentation must be requested from Ayonix where a filing is required.
Headquartered in Tokyo, with regional presence in Australia and the United States
What this does not establish
Regional presence, not a claim about the number of countries served or the location of deployments.
Shipped its first commercial face recognition SDK in 2014
ITOCHU Techno-Solutions Corporation adopted Ayonix FaceID across its offices in 2016 to replace RFID access cards, using iPads as facial-authentication terminals
What this does not establish
Deployment description only. No headcount, door count, error rate or cost saving is published.
Ayonix face recognition was deployed across more than 20 cameras at border gates for a United Nations border-control project
What this does not establish
Camera count and application only. No accuracy figure, throughput figure or country is published, because none has been released for publication.
6+ TOPS NPU inference
What this does not establish
Vendor-stated figure from the Ayonix product page. Awaiting confirmation against the hardware datasheet; listed in docs/APPROVALS-REQUIRED.md. TOPS is a peak arithmetic rating, not a face-recognition throughput figure, and no stream count is inferred from it.
4 TB local NVMe
What this does not establish
Vendor-stated. Retention duration depends on stream count and configuration.
2.5 GbE and 1 GbE dual networking
What this does not establish
Vendor-stated. Awaiting datasheet confirmation.
Integration work with Network Optix, Milestone and Genetec video management platforms
What this does not establish
Integration capability only. This is NOT a claim of certification, marketplace listing, technology-partner status or joint validation with any of the three vendors. No such status has been verified, and none is asserted anywhere on this site.
Founded by Dr Sadi Vural, who remains Chief Executive Officer
Doctorate from Osaka University, 2011, for the dissertation "Face recognition by using hybrid-holistic methods for outdoor surveillance systems"
The company name derives from the Ayofa-Filters introduced in the 2011 paper "Illumination Normalization for Outdoor Face Recognition by Using Ayofa-Filters"
ONVIF and RTSP camera input
What this does not establish
Protocol support. Whether a specific camera produces usable results depends on measured pixels across the face at the capture point, not on protocol compatibility.
On-premise, edge and fully air-gapped deployment
1:1 verification, 1:N identification, watchlist matching, face tracking, and liveness and presentation-attack detection
What this does not establish
Presentation-attack resistance is specific to the attack instruments tested for. Liveness is available in supported capture workflows, not universally.
Recognition is available to customer applications through an SDK and an HTTP API
What this does not establish
Confirms that an SDK and an HTTP API exist. It does not establish any specific function name, endpoint path, language binding or operating-system build, none of which is published on this site.
XProtect receives analytics events as XML on the Analytics Receive port, default 9090, or by HTTP POST, and matches the EventHeader Message against an alarm definition created in the Management Client
What this does not establish
Describes the documented Milestone protocol. It is not a statement of certification or marketplace status, and the port and prerequisites should be confirmed against the XProtect version in use.
The Security Center Web-based SDK role exposes Security Center SDK methods and objects as web services to support cross-platform development
What this does not establish
Quoted from the Genetec administrator guide. Full SDK documentation requires a Genetec Developer account obtained through the Development Acceleration Program, so version-specific capability is stated as unvalidated where it has not been tested.
The Nx Metadata SDK exchanges ObjectMetadataPacket and EventMetadataPacket with the Nx server, and the Rules Engine exposes an Analytics Object Detected event
What this does not establish
Describes the documented Nx plugin architecture. Not a statement of certification or marketplace listing.
ISO/IEC 30107-3 defines how presentation attack detection is tested and reported
What this does not establish
The standard describes a test methodology. It defines no pass mark and issues no certificate, so "ISO 30107-3 certified" describes nothing that exists. Ayonix makes no compliance claim against it; it is cited as the framework a buyer should require a test report to follow.
Rejected
These are recorded rather than quietly omitted, so the reasoning survives — including for claims that appear elsewhere and that someone may later ask why this site does not repeat.
Leading performance in Japanese immigration vendor tests associated with Narita and Haneda international airports
RejectedWhy: The parent site already labels this as Ayonix’s own report rather than an independently published result. A superlative about an unpublished test cannot be verified by a reader, so it is not published here in any form. It may be discussed under NDA in a procurement conversation.
Reviewed 2026-09-11 by Dr Sadi Vural, Chief Executive Officer. Source considered: ayonix.com/llms.txt, where it is disclosed as Ayonix’s own report ( link ) .
Any face recognition accuracy percentage, for Ayonix or any other vendor
RejectedWhy: A percentage without its threshold, dataset, gallery size and demographic breakdown cannot be reproduced, so it is not a fact a buyer can act on. Accuracy for a deployment is established by a pilot on the customer’s own cameras.
Reviewed 2026-09-11 by Jan Mocary, Chief Technology Officer. Source considered: Editorial policy, inherited from ayonix.com/technology/accuracy-and-liveness ( link ) .
Any count of customers, deployments, countries or installed cameras
RejectedWhy: Nothing in the researched Ayonix materials supports a count. Until Ayonix releases an auditable figure, no number appears on this site.
Reviewed 2026-09-11 by Gabriel Bamola, Chief Marketing Officer. Source considered: No verifiable source exists in Ayonix published materials .
Never published
Scanned for in the built HTML, in any casing. A page containing one of these does not ship.
Two of these deserve particular mention. "NIST certified" describes something that does not exist: NIST publishes comparative evaluations and does not certify, approve or endorse vendors, so the phrase is not a claim about Ayonix that happens to be false — it is not a claim about anything. "ISO 30107-3 certified" is the same error applied to presentation attack detection: the standard defines a test methodology, not a pass mark, and issues no certificate.
Because in a category where buyers have learned to discount vendor claims, the useful thing is not another claim but a mechanism. Every high-risk factual statement on this site is registered with its source, its source type, a verification date and a named reviewer before it may be published, and the build fails if a page states a claim it is not approved for. The register is the mechanism made visible.
Anything about accuracy, speed, scale, countries, customers, government deployments, awards, certifications, standards compliance, partner status, hardware specifications or platform support. These are the categories where vendors in this market most often state things that cannot be checked, so they are the categories this site refuses to state without a source.
The build fails. Claims are declared with the exact routes they are approved for, and the evidence components throw at build time if a page renders one it is not cleared for. The quality gate additionally scans the built HTML for banned phrases and for high-risk claim language on pages that are not approved for it. This is enforcement rather than policy.
Because they were reviewed and found unsupportable, and recording the reason is more useful than silently omitting them. A rejected claim stays in the register so that the reasoning survives — including for claims that appear elsewhere and that someone may later ask why this site does not repeat.
Write to infojp@ayonix.com with the page, the statement and the correction. Corrections are made in public: when the text changes, the last-reviewed date on the page changes with it. We would rather be described accurately than favourably, and that applies to our own site as much as to anyone else’s description of us.
Corrections
Write with the page, the statement and the correction. Corrections are made in public and change the page’s last-reviewed date. We would rather be described accurately than favourably.