Skip to main content
Cameras and WiFi CSI solve overlapping but not identical problems. The choice is about what data you need from the sensor, not just privacy preferences.

​
Sensing mechanism

A camera answers “what does this look like?” CSI answers “how did the radio channel change?”

​
Coverage geometry

Cameras need line of sight per zone: one camera per sightline, blind behind furniture and walls. CSI propagates through drywall and around obstacles; a node pair can cover a whole room without aiming. Cameras win when you need per-pixel detail in a specific field of view. CSI wins for whole-room, NLOS, darkness-invariant occupancy and motion.

​
Latency and compute

Camera pipelines run capture, decode frame, run vision model, then emit event. Typical end-to-end latency is 50 to 200 ms on edge hardware, higher if frames go to cloud. CSI pipelines run capture packet, preprocess, feature extract, then infer. Latency is comparable on the same host, but the input is orders of magnitude smaller: kilobytes of complex numbers versus megapixel frames.

​
Deployment cost

A room with four sightlines needs four cameras, cabling, storage, and often cloud processing. The same room needs two ESP32 nodes and a host running the Python pipeline. No video storage, no bandwidth for frame upload.

​
When to choose which

Camera: identity verification, reading text, precise pose in a controlled FOV, forensic video. WiFi CSI: anonymous occupancy, motion in the dark, whole-room presence, energy analytics, elder-care awareness without surveillance. See occupancy detection and the detection ladder for what CSI delivers without images.