Face Recognition for Existing CCTV
Face recognition for people on the move.
AI Bot Eye recognizes enrolled people as they walk through suitable CCTV views. Use live camera feeds or recorded NVR video, with recognition processing on site and no internet required for the core system.
Live feeds or recordings. Camera suitability checked before deployment.
Face Recognition?
AI Bot Eye is an on-site face recognition system that compares faces captured from live CCTV or recorded NVR video with enrolled identities. It is designed for people walking through suitable camera views, and its core recognition processing can operate without internet connectivity.
See the system
Start with a demonstration.
Then test your own view.
Watch the AI Bot Eye Face Recognition demo, then bring us the entrance, passage or recording you want to assess. Your camera view is the next step in understanding whether the system fits your site.
During your test, ask to see recognition events, the operator interface and the enrolment process.
Designed around movement
People keep walking.
The camera captures the opportunity.
Many attendance terminals ask someone to stop and present their face. AI Bot Eye is designed to recognize people walking through monitored areas, where a suitable camera can capture a usable face view.
A deliberate check-in
Someone presents their face at a dedicated station.
The interaction is usually centred on an attendance or entry transaction.
The person can usually see whether the check-in succeeded.
A recognition event in the flow
People walk through a suitable monitored passage.
The camera supplies face observations for comparison with enrolled identities.
The relevant event becomes available to the operator or configured alert workflow.
Recognition still needs a usable view of the face. Camera placement and real site conditions matter.
Two ways to work
Recognize now.
Review what was recorded.
As people pass the camera
Process connected live feeds and compare captured faces with enrolled identities. A watchlist match can create a recognition event and trigger the configured alert workflow.
From footage already captured
Run recognition on previously recorded video to review appearances. The recording must contain usable face detail; processing an old clip cannot recover detail the camera never captured.
From recognition to awareness
Give the right people
a reason to pay attention.
Configure lists around your operational purpose. Review matches in context and decide what the team should do next.
A more informed welcome
Assess recognition at the lobby or reception entrance for enrolled VIP guests. Discuss which service team should be notified and demonstrate that workflow during the test.
Separately, a hotel watchlist can help security review the arrival of a banned person or vendor.
Focus on relevant arrivals
Use an approved-person list to suppress alerts for enrolled staff where alerts are unnecessary. Maintain a watchlist for identities that require security attention.
Operators manage the identities and lists that matter to the passage.
Review the entrance event
Assess recognition at selected entry points and review captured events. Process suitable recorded footage when the team needs to review previous appearances.
Start with selected views and a defined response, then evaluate expansion.
A face match is information for review. An unrecognized person is not automatically an unauthorized person, and recognition alone does not establish a right to enter. For zone and after-hours movement alerts without identity matching, see intrusion detection on existing CCTV.
Practical enrolment
Start with a photo.
Build a better reference.
Use an existing photograph
Add a person using an available face image, including a suitable passport-style or official photograph. Begin with a clear reference that represents the person you want to enrol.
Prefer four to six views
For better reference coverage, provide four to six photographs from different angles, including a clear frontal image. Reference quality and camera conditions work together.
Review camera-captured identities
The system can initially create provisional identities from observed faces. An operator can review and rename an identity, then manage its list membership. The camera does not discover the person’s name by itself.
Face search
Have a photo.
Find matching faces.
Upload a clear face image and search against people already enrolled on site. Operators can review the closest matches and decide what to do next.
Start with a face image
Use a photo where the face is visible enough to compare — a still from CCTV, an ID-style image, or another clear reference.
Match against enrolled people
The system compares the uploaded face with enrolled identities in your local database. Processing stays on site with the rest of the recognition workflow.
Inspect the closest matches
Review returned matches before acting. A search result is a candidate for operator review, not an automatic identity decision.
Useful when security or facilities teams already have a photo and need to check whether that person is enrolled — or which enrolled people look similar.
The view matters
A usable face view
starts with camera placement.
A reasonably well-lit passage with a more centred, frontal view is a strong starting point. Assess real movement and lighting across the day.
Make the face visible
Camera resolution, physical distance, height and angle affect how much useful face detail is captured. A camera mounted too high can produce a steep view; a camera at the end of a wide passage may miss a centred face as people walk off-axis.
Check the light
Strong backlighting can leave faces dark. Semi-open entrances need checks from morning through evening as exposure changes.
Check the movement
Side profiles can work, but a more frontal view is preferable. Test normal walking and the actual flow through the passage.
Local by design
Recognition stays
inside your site.
All recognition processing happens on site. Live feeds and recordings follow the local processing path shown above; nothing needs to be uploaded elsewhere for core recognition.
Control-room operators can manage people and enrolled faces. External notification services may require connectivity even when recognition itself operates offline.
Intended current device configuration. Stream and identity capacity depend on workload and site assessment; these are not universal performance guarantees.
Experience behind the system
Built through research.
Informed by real deployments.
Rao IT’s face recognition work began during the founder’s Ph.D. research in 2019–2020. Subsequent projects brought practical learning about camera resolution, distance, light and changing exposure.
These examples describe the wider technology history, not a list of features bundled with this CCTV configuration.
CCTV and classroom attendance
A college attendance project explored CCTV recognition. Another workflow used three to five classroom photographs to mark attendance for approximately 50–65 students.
Finding guests in event photographs
At an annual event with approximately 400 guests, face matching helped return relevant candid and group photographs in response to a guest’s selfie.
Examination authorization
Face recognition has been used to authorize exam-paper downloading for printers at examination centres, across government, university and state-level examination projects.
Your view is the next test
Start with the cameras
that matter to your team.
We can test on your live CCTV feeds or recorded footage. Begin with a defined use case and selected views, then review what the results mean for deployment.
Discuss the use case
Identify the people, passage and operational response you want to assess.
Assess camera views
Review angle, distance, light and enrolment references together.
Test live or recorded
Demonstrate recognition with your feeds or recordings and review the events.
Review rollout fit
Agree on adjustments and suitability before expanding to more views.
Bring a representative view
Choose footage that reflects the conditions your team actually works with: normal walking speed, typical passage traffic and the lighting at relevant times. Include the enrolled reference images you intend to use. A carefully selected demonstration frame alone cannot answer whether the everyday view is suitable.
Review the complete workflow
Ask the team to show a captured event, how an operator manages the associated identity and what happens for the configured list. Review missed or uncertain observations alongside successful recognition. Agree which camera adjustments and reference images should be tested again before deciding on deployment.
For security teams, the test should clarify which events deserve attention. For facilities teams, it should clarify camera placement and practical changes. For IT teams, it should clarify local processing, feed availability and any connectivity needed by the chosen notification workflow.
Practical answers
Before you connect
your first camera.
The essentials for security, facilities and IT teams assessing CCTV face recognition.
Can it recognize people while they walk?
Yes. AI Bot Eye is designed for people walking naturally through suitable CCTV views. Recognition depends on usable face detail, lighting, angle, distance and enrolment quality.
Can it work with our existing CCTV?
Existing CCTV can be assessed for suitability. We review the available feeds and camera views before deployment; compatibility and useful face detail should be confirmed in a test. A camera that is useful for general surveillance may still need repositioning to capture a clearer face. Start with the specific entrances and passages where recognition would help your team.
Can it process recorded NVR video?
Yes. The system can process previously recorded NVR video as well as live CCTV. This does not mean every recording is automatically indexed or instantly searchable.
Does it require internet or upload footage?
Core recognition processing happens on site and can operate without internet. It does not require footage to be uploaded elsewhere. External notification services may have separate connectivity requirements.
Can I search with a photo?
Yes. Face search lets an operator upload a clear face image and compare it with enrolled people on site. Review the closest matches before acting — a search result is a candidate for review, not an automatic identity decision.
How are people enrolled?
Use an existing photograph, preferably four to six reference images from different angles, or let an operator review and rename a provisional identity created from observed faces.
Can it identify someone whose name is unknown?
It can create a provisional face identity for operator review. That does not reveal a person’s name. A named recognition result requires an identity associated with the enrolled reference.
How do watchlists and approved-person lists work?
A recognized watchlisted identity can trigger a configured alert workflow. An approved-person list can suppress alerts for people such as enrolled staff. Operators manage these lists for the intended purpose.
What lighting and camera placement are needed?
Start with a reasonably well-lit passage and a clear, more frontal view. Avoid strong backlighting and overly steep camera angles. Assess exposure changes and normal movement in the actual location.
How many streams and enrolled faces can it support?
The intended current device configuration is approximately 10 live CCTV streams and 300 enrolled faces. Capacity is subject to workload and site assessment.
Can we test using our own footage?
Yes. Request a test on your live CCTV feeds or recordings. The team will discuss the use case, assess the views and review recognition results with you. The first conversation can establish which footage and enrolled references are useful, how to make them available for the test and what your operators need to see before a rollout decision.
How is this different from an attendance terminal?
A typical terminal involves deliberately presenting a face for a transaction. AI Bot Eye processes suitable CCTV views as people move through them, creating recognition events for the configured operational purpose.
Prove the fit on your site
Your cameras.
Your conditions.
A test you can review.
Bring a live feed or a recording. Start with one useful view.