Hikvision DS-2CD2047G3-LI2UY – 4 MP Smart Hybrid Light ColorVu Bullet

Both datasheets are public, so this is a comparison rather than a claim.

One oddity worth knowing before you design around it. Hikvision publishes identical DORI figures for both cameras — 61 m detect, 24 m observe, 12 m recognise, 6 m identify at 2.8 mm — even though the G3 spreads the same four megapixels across seven more degrees of scene. A wider lens normally costs recognition distance. Both sets are the manufacturer’s own published numbers and both are reproduced here as published; we have not recalculated them, and neither should a specification that has to stand up later.

The Smart Hybrid Light works on the principle the name describes: infrared while the scene is quiet, white light when the camera classifies a person or a vehicle, so the light is triggered by deep-learning classification rather than by motion. The G3 datasheet lists three supplement-light modes and does not name them, where the previous generation documented four. If a customer needs the exact mode behaviour written into a specification, confirm it against current Hikvision documentation — the mechanism is explained at length on the DS-2CD2047G2H-LIU page, but the mode list there belongs to that model.

The lens is fixed and chosen at order. Switch between the two below: the 4 mm trades sixteen degrees of width for fourteen metres of detection range.

Read the result against the 40 m supplement light. At either focal length the detection distance runs past what the light reaches, so on an unlit site 40 m is the honest working edge after dark. Every angle and distance here is the manufacturer’s.

The camera makes exactly one decision on its own: is that shape a person or a vehicle. That decision is what turns the white light on, and it is genuinely useful. It is also the entire extent of what the camera understands. It cannot tell a delivery driver from a stranger, and it does not know that the same figure at three in the afternoon and three in the morning mean different things.

SmartGard.AI needs one thing from a camera: the ability to say that something happened. This one reports motion with human and vehicle classification, line crossing, intrusion, region entrance and exiting, face capture, scene change and tampering. When a trigger arrives, the frames from that event are read against zone rules written in plain language, and the system decides whether anybody needs to hear about it.

The hybrid light matters here for a practical reason: the frame that arrives is in colour rather than infrared grey, which is a better thing to reason over. Clothing, vehicle colour and detail survive.

Two honest limits. Each event is judged on its own — there is no correlation between cameras, so a figure crossing three fields of view is three separate judgements, not one tracked path. And nothing is watched continuously: analysis runs when an event fires or when a patrol is requested from the app. This is a layer that decides what deserves your attention, not a monitored alarm system with a receiving centre behind it.

Already have cameras up? The same layer retrofits onto most modern IP installations without replacing what you already own.