Marketing vs Reality: An Honest Assessment of RuView's Claims

[lanceharvie]

The core misrepresentation

RuView markets itself as turning "ordinary WiFi" into spatial intelligence. This is misleading. Your home router is just an RF illuminator. The actual sensing requires a purpose-built wireless sensor network of dedicated ESP32-S3 or Raspberry Pi nodes running custom firmware, physically deployed throughout the space. That is a sensor network, not a WiFi feature.

No different from cameras at the infrastructure level

The approach requires the same fundamental commitment as a camera network: multiple nodes placed for overlapping coverage, wiring or battery power for each, configuration per node, and ongoing calibration. The claim that this avoids infrastructure complexity is false. A $30 PIR sensor or a $100 smart camera is vastly simpler to deploy for 95% of home automation use cases.

Accuracy claims do not hold up

The headline feature 17-keypoint pose estimation is admitted in the beta notice to run at 2.5% PCK@20. MediaPipe on a standard webcam runs at 95%+. These are not comparable. Presence detection and vital signs are the mature parts of the stack. Pose estimation is experimental at best.

The consumer home use case is the most overstated

Flashing custom firmware, provisioning nodes, tuning for room geometry, handling interference from real WiFi traffic and neighbouring networks, and recalibrating when furniture moves is not a consumer experience. This requires embedded systems expertise to operate reliably.

The Cognitum Seed is the tell

The existence of a commercial appliance product implicitly acknowledges that the open source stack is not deployable without significant technical effort. The repo functions as marketing for hardware that has not yet shipped.

What this actually is

A legitimate and technically impressive research project with a real niche: privacy-constrained environments where cameras are legally or ethically off the table, bedrooms, bathrooms, medical monitoring. Outside that niche, cameras plus computer vision win on every performance metric. The 63k stars reflect enthusiasm for the concept, not production deployments.

[Claori]

The only thing I would mention, cameras don't win when it comes to Epilepsy, but for Epilepsy/Medical monitoring....I would recommend transforming the CSI to CIR....as CIR is much, much more superior in regards to medical sensing.

Feature	Wi-Fi CSI (Channel State Information)	UWB CIR (Channel Impulse Response)
Spatial Resolution	Low to Moderate. Groups data across wide subcarriers.	Extremely High. Pinpoints micromovements down to millimeters.
Multipath Interference	High. Signal easily gets distorted by moving pets, fans, or shifting furniture.	Very Low. Isolates specific signal reflections in time, ignoring background clutter.
Micro-Movement Tracking	Struggling. Hard to isolate a rapid, subtle tremor from basic breathing.	Excellent. Easily separates a tonic-clonic muscle spasm from a person rolling over.
Privacy & Form Factor	No lenses (excellent privacy), but requires bulky routers/receivers.	No lenses (excellent privacy), using tiny, low-power sensor nodes.