Do we need to replace our CCTV cameras to use AI video analytics?
No, in almost all cases. AKCS Vision reads your existing IP cameras over RTSP, and all processing runs on hardware installed inside your property.
Do we need to replace our CCTV cameras to use AI video analytics?
No, in almost all cases. AKCS Vision connects to the IP cameras you already have, over their RTSP stream. Most CCTV installed in Indian communities in the last several years exposes an RTSP stream, and that is all Vision needs.
Brand and installer do not matter. What matters is that each camera hands over a live RTSP stream on your local network, and that it is pointed at something worth analysing. The cameras that genuinely will not work are old analogue ones on a DVR with no IP output at all.
Where does the video actually get processed?
On hardware installed inside your property. AKCS supplies and installs an on-premise Vision box in your server room. It pulls the camera streams over your local network, runs the detection models there, and sends only the resulting events to the AKCS platform.
The practical consequence is that your video does not leave the premises. What travels out is an event record — what was detected, on which camera, at what time — not a continuous video feed. Your existing NVR keeps doing whatever it already does with recordings.
AKCS deploys and manages that on-site hardware as part of rollout. It is not something your committee has to specify, buy, or maintain.
What does AKCS Vision detect?
Ten detection types. This is the complete list — Vision does not have a hidden catalogue of extras, and nothing outside this list is being detected.
| Detection | What it catches |
|---|---|
| Person fall | Someone collapses or falls to the ground and stays down |
| People count exceeded | More people in an area than the count you set for it |
| Crowd anomaly | A crowd forming or behaving unusually for that spot |
| Motion at unusual times | Movement in an area during hours you marked as quiet |
| Loitering | A person staying in one place far longer than normal |
| Intrusion | Someone entering a zone you marked as off-limits |
| Violence | Physical fighting or an assault |
| Fire | Visible flame |
| Smoke | Visible smoke |
| Abandoned object | A bag or parcel left unattended in a public area |
Each one is configured per camera, because a rule that makes sense at the basement ramp makes no sense at the clubhouse. Loitering at the gate at 2am is worth an event. Loitering near the children's play area at 6pm is just an evening.
What happens when the system detects something?
The event is created with status new and appears in the Vision events list for your management team. From there a person moves it through the workflow. Nothing is auto-closed and nothing is auto-escalated to residents.
| Status | Meaning |
|---|---|
| New | Just detected, nobody has looked at it |
| Acknowledged | Someone has seen it and taken ownership |
| Investigating | Being checked — footage pulled, security personnel sent |
| Resolved | Dealt with, and the outcome recorded |
| False positive | The detection was wrong, and it is marked as such |
Marking false positives honestly matters. It is how your committee finds out that a particular camera is throwing noise because of a swaying tree or a headlight glare, and that camera can then be re-tuned or re-aimed.
Does it read number plates?
Yes, at the gate, and the number-plate recognition runs on the Vision hardware on your premises like every other detection. It does not depend on a cloud service to read a plate.
What does the community have to provide?
A small server room, and network reach to your cameras. That is the whole list on your side.
- A lockable room. The Vision box and your existing NVR should sit behind a door only your management team can open.
- Cooling. The room needs to stay cool. A hot cupboard shortens the life of every piece of equipment in it, including the recorder you already own.
- Network. The Vision box must reach your cameras on the local network, and reach the internet to post events.
- Power. A stable supply, ideally on the same backup that keeps your CCTV running through a cut.
If your CCTV already sits in a proper room today, you very likely need to change nothing.
What decides whether a camera works well?
Placement, height, angle, night lighting, and the state of the lens. AI does not rescue a badly installed camera. A camera that produces a poor picture for a human produces a poor picture for a model.
- Placement — a camera watching a wall detects nothing worth having.
- Height and angle — mounted too high and pointed too steeply down, you get the tops of heads. Falls, violence and loitering all need to see a body, not a scalp.
- Night lighting — this is the most common weakness. A corner that is unusably dark at 1am will not produce reliable detections at 1am.
- Lens condition — dust, spider webs, water marks and haze on an old dome cover degrade everything downstream.
How do we find out if our cameras will work?
With a free hardware audit, before any quote. AKCS surveys your existing cameras, checks which ones expose a usable RTSP stream, and reports which positions are good, which need re-aiming, and which genuinely need replacement.
Re-aiming three or four cameras is cheap. Replacing a whole tower of them is not, and it is usually unnecessary. You get the assessment before any pricing conversation, so see pricing, then book a demo to arrange the audit.
Still stuck?
Email contact@shinraidynamics.com and a human replies within the same working day.