Edge Computing Fundamentals for CV
What You'll Learn
You'll learn when and why to deploy computer vision models at the edge versus the cloud, and understand the critical constraints that shape those decisions.
The Edge vs. Cloud Decision
Edge computing means running CV models directly on devices near where data is captured—security cameras, drones, smartphones, IoT sensors—rather than sending data to remote cloud servers.
Think of it like cooking at home versus ordering delivery. Cloud is the restaurant: powerful kitchens (GPUs), but you wait for delivery (latency). Edge is your home kitchen: faster meals (instant response), but limited equipment (constrained hardware).
Four Key Constraints
Latency
Edge deployment eliminates network round-trips. For autonomous vehicles detecting pedestrians or AR apps tracking faces, milliseconds matter. Cloud latency (50-200ms+) may be unacceptable; edge inference (<10ms) becomes necessary.
Power
Edge devices often run on batteries or limited power budgets. A surveillance camera can't consume 300W like a cloud GPU. You'll need quantized, pruned models that balance accuracy with efficiency.
Bandwidth
Streaming raw 4K video to the cloud costs money and may be impossible in remote areas. Processing locally means sending only results (e.g., "person detected" vs. 30fps video stream).
Privacy & Security
Processing medical images or home security footage locally keeps sensitive data on-device, avoiding transmission risks and regulatory issues.
When to Choose Edge
Deploy at edge when you need: real-time response, limited connectivity, privacy compliance, or cost-effective scale (thousands of devices).
Use cloud when you need: maximum accuracy, complex multi-model orchestration, or centralized training data collection.
Key Takeaway: Edge deployment trades computational resources for lower latency, reduced bandwidth, and enhanced privacy—choose based on your application's bottleneck, not default assumptions.