Camera and deployment requirements
What A-Eye needs from your cameras, your network and your site. Written for the person who will actually have to make it work.
Standard IP cameras exposing an RTSP stream, an edge device on the same local network, and a view of the thing you want verified. Placement matters more than camera specification in most deployments.
What actually matters, and what does not.
| Attribute | Workable | Notes |
|---|---|---|
| Stream protocol | RTSP, ONVIF compatible | Most IP cameras and NVR or encoder outputs qualify. Analogue via encoder is usually fine. |
| Resolution | 1080p is sufficient for most use cases | Pixels on the target matter, not sensor size. A tighter 1080p view beats a wide 4K view. |
| Frame rate | 10 to 15 fps is enough for most verification | High frame rate matters only for fast moving objects. Higher fps costs processing capacity. |
| Codec | H.264 or H.265 | Both are standard and supported. |
| Low light | Infrared or consistent artificial lighting | An area that goes effectively dark cannot be verified without infrared. |
| Mounting | Fixed position | PTZ cameras that move break zone definitions unless parked on a preset. |
If a camera already gives a human a clear view of the activity, it is usually adequate. The most common change we recommend is angle, not equipment.
Camera position by question.
| What you want verified | Position that works | Common mistake |
|---|---|---|
| Production count | Across the line, perpendicular to travel | Looking along the line, so items occlude each other |
| PPE compliance | Facing the approach at head and torso height | Ceiling camera looking straight down, where a helmet is barely visible |
| Pallet and stock counting | Down the aisle with racking in frame | Wide overview covering several aisles at once |
| Forklift and pedestrian proximity | Elevated corner view covering the junction | A view that only captures one approach |
| Dock and dispatch | Covering the bay and the approach lane | Camera positioned inside the bay only |
| Number plates | Dedicated camera at plate height, near head on | Reusing a general area camera at a steep angle |
What sits on site.
Edge device
An NVIDIA Jetson class device on the same local network as the cameras. It pulls streams, runs detection and verification locally, and stores events and evidence frames. Video is not sent to a cloud service for processing.
Network
The device needs to reach the cameras on the local network. It does not need the camera network exposed to the internet. Outbound connectivity is used for the console and for pushing verified events to your ERP, not for streaming video out.
Power and placement
Standard mains power in a comms room or cabinet. Industrial environments should account for temperature and dust in line with normal IT equipment practice.
Sizing
Camera count per device depends on resolution, frame rate and how many verification types run per stream. Our validated envelope is up to roughly ten cameras per device. We size it against your actual streams rather than a headline figure.
Pre deployment checklist.
Gather
- Camera list with model, resolution and location
- A sample recording from each camera you want used
- Site or floor plan, even a rough one
- The specific questions you want answered
- Which system should receive the verified events
Confirm
- Cameras expose an RTSP stream and credentials are available
- Positions are fixed, or PTZ cameras sit on a preset
- Areas of interest are lit consistently across the shift
- Network access from the edge device to the cameras
- Who internally owns acting on the exceptions raised
The last item is the one most often skipped and the one that most often decides whether a deployment produces value. Verification only pays back if somebody acts on what it surfaces.
Common questions
Do we need to buy new cameras?
Usually not. Most deployments run on cameras already installed. Where a specific question cannot be answered from the current view, the fix is more often repositioning an existing camera than buying a new one.
What if our cameras are on a closed network?
That is normal and generally preferred. The edge device sits on the same network and processes streams locally. It does not need the camera network to reach the internet.
How much bandwidth does it use?
Very little on the wide area network, because video is processed on site and is not uploaded. What crosses the network is the camera stream between camera and edge device, which stays on the local network.
Does it record video as well?
A-Eye retains frames as evidence for verified events. It is not a replacement for a recorder or VMS, and most sites keep their existing recording in place.
How many cameras can one edge device handle?
It depends on resolution, frame rate and how many verification types run per stream. Our current validated envelope is up to roughly ten cameras per device, and we would rather size that honestly during assessment than quote a number that does not hold on your site.
Find out what your cameras can actually verify
An operational assessment looks at your cameras, your zones and your questions, then tells you which are answerable today and which are not.