12 product modules built around safety analytics, crowd flow, anomaly detection and post-event investigation — grounded in academic research in gait and human movement, with PrivacyGuard defining the default design posture and auditability by design.
From gait identity and movement patterns to unusual behaviour and safety events — SecureVision turns human-motion signals into privacy-aware, actionable insight.
What SecureVision does, in the words security and operations teams use for it — and which modules and environments each capability lives in.
Compare the source, its estimated pose and a feature-level explanation. Reducing appearance detail can support privacy-aware workflows; pose and movement may still carry identifying information.
Switch the view
Each group carries its own governance requirements, and the two are never interchangeable.
Anomaly detection, crowd flow, worker safety and campus monitoring run on movement features rather than identity. PrivacyGuard is designed to support skeleton-only processing, optional face blur, role-based access and configurable retention at the pipeline level. This is the default posture for SecureVision — privacy-aware architecture, not a guarantee of anonymity.
Surfaces candidate movement events — loitering, running, restricted-zone entry, tailgating-like patterns and perimeter events — for operator review, without requiring identity recognition.
Analyzes crowd movement, density, queues, bottlenecks, flow direction and abnormal crowd-motion indicators for smart-city scale public spaces.
Flags movement patterns associated with falls, slips, restricted-zone entry, fatigue-like movement and emergency movement events across industrial sites.
The architectural privacy layer for GaitAI: designed to support skeleton-only analytics, face blur, role-based access, configurable retention and audit logs in privacy-sensitive environments.
Surfaces movement-safety events across office parks, universities, hospitals and corporate campuses for operator review.
Provides crowd-movement indicators for high-density public events: stadiums, conferences, religious gatherings, concerts, rallies, exhibitions.
Surfaces unusual movement, loitering, queue congestion, staff-safety and emergency-flow indicators inside retail environments for operator review.
Facility-safety and authorised-access movement intelligence configured for defence installations, in Army, Navy and Air Force modes. Human-supervised throughout: an indicator supplements existing procedure and never acts on its own.
Cross-camera correspondence, gait-assisted access control, post-event investigation and policy-governed watchlist matching. These produce confidence-based candidates for trained review — never proof of identity — and they are not offered for general-public surveillance.
Identity-related capabilities are intended only for lawful, authorized environments with appropriate access controls, governance and auditability. Where movement without identity matching intelligence is sufficient for the outcome, that is the intended default.
Searches uploaded CCTV footage for a person, movement pattern, timeline, or event path after an incident.
Produces confidence-based candidate matches across multiple camera feeds using movement and body-level signatures rather than face-only matching. Candidates are for trained review — never proof of identity.
Contributes a passive gait-consistency signal alongside card, face or mobile authentication in high-security spaces. It supports an existing credential rather than replacing it.
Policy-governed candidate matching against an authorized list, with confidence scoring, access controls and audit history. Restricted to environments with lawful authority — not offered for general-public surveillance.
Explore the SecureVision app
GaitAI SecureVision is a native Android app for spaces you are authorised to analyse. Upload a short video and an on-device person detector counts who it can see, frame by frame, or draw a zone and track entries, occupancy and dwell. It detects people as boxes and identifies no one. These are the app's real screens.
Free download. An active subscription is required to use the application's features. Plans and payment options are available inside the app.
1 of 9 · HomeCrowdSense is recommended first. Your most recent zone analysis and the rest of the catalogue sit underneath.
Coming soon, visible but not startable: Motion analysis (where and when movement happens).
Version 0.2.0 (build 3) is a public release, free to download and signed with the GaitAI release key. Sign in with a GaitAI account inside the app; analysis, history and reports unlock with an active subscription, chosen and paid for in the app.
Version 0.2.0 (build 3) · Android 7.0 or later · free download.
Counts are people visible to the detector in each sampled frame, not a crowd-density estimate; entries, exits and dwell are tracker estimates and are labelled so. Only analyse footage you own or are permitted to analyse.
PrivacyGuard is the architectural privacy-control layer for SecureVision, designed to support skeleton-only analytics, face blur, privacy-aware aggregated heatmaps, role-based access, audit logs, retention controls and consent/policy logging. What a given site enables is a deployment decision.
PrivacyGuard is privacy-aware architecture — it minimises identifiable data and governs access. It is not a guarantee of anonymity, and no compliance certification is claimed.
Separate from the processing path below — these are controls over who may see what the last step produces.
Three steps, one figure. Step through them and see what each one carries forward — and what stops being available.
What a camera holds before anything has been done to it.
This path describes the privacy-aware group — anomaly detection, crowd flow, worker safety, campus and access monitoring. SecureVision's identity and investigation modules are a separate, governed group that does retain identity-bearing information by design, and they do not run this path.
An illustrative architecture diagram, not a product specification or a measurement. It shows minimisation and transformation — not anonymity: pose and gait can themselves carry identifying information, which is why GaitAI researches the question rather than claiming it away.
SecureVision is built for a range of operational environments — each with its own problem, product mix and output.
Crowded transport hubs need anomaly + flow intelligence — without requiring identity matching for flow analysis.
Public spaces need crowd intelligence designed for real-time operational workflows, with privacy respected by default.
Campuses need quiet, privacy-aware safety monitoring across many sites.
Worker falls and unsafe-zone entries can be difficult to spot in time.
Loss-prevention, queue management and staff safety on one floor — with appearance-reduced movement analysis.
High-density events need crowd-movement awareness during the event, not after the news cycle.
Defence and military organisations need facility-safety and access intelligence for personnel across very different installations — without identity-first monitoring of the people they have a duty of care to.
