What edge computing means
Edge computing means running the model directly where the data is produced — in a camera, a machine or a mobile device — rather than sending images to the cloud for analysis. This matters when an immediate response is needed, when the network connection is unreliable, or when images must not leave the device for privacy reasons. In computer vision, edge computing is often a natural choice for exactly these reasons.
The advantages
The three biggest advantages of edge computing are speed, independence from the network and privacy. A response arrives in milliseconds because the image does not need to be sent anywhere. The device keeps working even when the connection drops. And because images never leave the device, sensitive material stays local. This combination is decisive in many industrial and healthcare applications.
Limitations and trade-offs
Edge devices have limited compute and memory, so the model must be lightweight. This often means a large cloud model is slimmed down, for example by compression or pruning, which can slightly lower accuracy. The trade-off must be found case by case: how much accuracy can be sacrificed for speed and locality. When well designed, the difference is often negligibly small.
The hybrid model
Often the best solution is a hybrid: a lightweight model on the device handles the fast, common cases, and uncertain or demanding cases are sent to the cloud for more thorough analysis. This brings the speed of the edge and the accuracy of the cloud into one system without having to sacrifice either.
Summary
Computer vision at the edge is worthwhile when you need an immediate response, operation without a network or strict privacy. The choice requires a lighter model and careful optimisation, but done right it brings speed and privacy that the cloud cannot offer. Often a hybrid solution gives the best of both worlds.