Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ Works natively with the four major smart-home ecosystems: **[Home Assistant](doc

[![Works with Home Assistant](https://img.shields.io/badge/Works%20with-Home%20Assistant-blue?logo=home-assistant&logoColor=white&labelColor=41BDF5)](docs/integrations/home-assistant.md) [![Works with Matter](https://img.shields.io/badge/Works%20with-Matter-blue?labelColor=4285F4)](docs/adr/ADR-122-bfld-ruview-ha-matter-exposure.md) [![Works with Apple Home](https://img.shields.io/badge/Works%20with-Apple%20Home-black?logo=apple)](docs/user-guide-apple-homepod.md) [![Works with Google Home](https://img.shields.io/badge/Works%20with-Google%20Home-blue?logo=googlehome)](docs/integrations/home-assistant.md) [![Works with Alexa](https://img.shields.io/badge/Works%20with-Alexa-blue?logo=amazon&logoColor=white&labelColor=00CAFF)](docs/integrations/home-assistant.md)

> Drop into any **Home Assistant** install with one `--mqtt` flag. Or pair into **Apple Home / Google Home / Alexa / SmartThings** as a Matter Bridge. Ships 21 entities per node (11 raw signals + 10 inferred semantic states: someone-sleeping, possible-distress, room-active, elderly-inactivity-anomaly, meeting-in-progress, bathroom-occupied, fall-risk-elevated, bed-exit, no-movement, multi-room-transition) plus 3 starter HA Blueprints. See [`docs/integrations/home-assistant.md`](docs/integrations/home-assistant.md) · [ADR-115](docs/adr/ADR-115-home-assistant-integration.md).
> Drop into any **Home Assistant** install with one `--mqtt` flag. Or pair into **Apple Home / Google Home / Alexa / SmartThings** as a Matter Bridge. Ships 15 entities per node (9 raw signals + 6 inferred semantic states: someone-sleeping, possible-distress, room-active, elderly-inactivity-anomaly, fall-risk-elevated, no-movement) plus 3 starter HA Blueprints. See [`docs/integrations/home-assistant.md`](docs/integrations/home-assistant.md) · [ADR-115](docs/adr/ADR-115-home-assistant-integration.md).

### π RuView is a WiFi sensing platform that turns radio signals into spatial intelligence.

Expand Down Expand Up @@ -335,7 +335,7 @@ levels. Read the label, not the headline ([ADR-187](docs/adr/ADR-187-archive-v1-

| Tier | Checkpoint(s) | Honest status |
|------|---------------|---------------|
| **Real & validated** | [`ruvnet/wifi-densepose-pretrained`](https://huggingface.co/ruvnet/wifi-densepose-pretrained) (CSI encoder + presence head) · [`ruvnet/wifi-densepose-mmfi-pose`](https://huggingface.co/ruvnet/wifi-densepose-mmfi-pose) (17-keypoint pose) · `cog-person-count/count_v1` | **MEASURED / published.** Presence = 82.3% held-out temporal-triplet accuracy (the old "100% presence" figure was retracted); MM-Fi pose = 82.69% torso-PCK@20 on the `random_split` protocol. These are the pose/presence numbers the project stands behind today. |
| **Real & validated** | [`ruvnet/wifi-densepose-pretrained`](https://huggingface.co/ruvnet/wifi-densepose-pretrained) (CSI encoder + presence head) · [`ruvnet/wifi-densepose-mmfi-pose`](https://huggingface.co/ruvnet/wifi-densepose-mmfi-pose) (17-keypoint pose) · `cog-person-count/count_v1` | **MEASURED / published.** The CSI encoder scores 82.3% held-out temporal-triplet accuracy (the old "100% presence" figure was retracted) — that is an embedding-quality metric, not a presence decision rate. The published presence head on top of it is degenerate: its bias exceeds its weight norm (8.19 > 3.67), so on L2-normalised embeddings every logit lands in `[4.52, 11.86]` and it cannot output `absent` for any input ([#1521](https://github.com/ruvnet/RuView/issues/1521)); the [ADR-298](docs/adr/ADR-298-model-release-sanity-gates.md) release gate rejects it. MM-Fi pose = 82.69% torso-PCK@20 on the `random_split` protocol. |
| **Real but weak (honestly labeled)** | committed `v2/crates/cog-pose-estimation/cog/artifacts/pose_v1.safetensors` | First-cut on-device model. **PCK@20 = 3.0% / PCK@50 = 18.5%** on a 217-sample holdout — **below the ADR-079 target of ≥ 35%.** Learns coarse structure (`r_hip` 77% PCK@50); distal/face joints near-random. Its runtime path in `cog-pose-estimation/src/inference.rs` is still a centred-skeleton **stub returning `confidence=0`** — the weights are not yet wired in. Full disclosure in the [cog README](v2/crates/cog-pose-estimation/cog/README.md). |
| **Architecture only, no weights** | `archive/v1` `DensePoseHead` | Random `kaiming_normal_` init, **no checkpoint of any kind** (zero `.pth`/`.onnx`/`.safetensors` files under `archive/v1/`). Deprecated and superseded — see [`archive/v1/DEPRECATED.md`](archive/v1/DEPRECATED.md). Do not expect real pose output from it. |

Expand Down
8 changes: 7 additions & 1 deletion docs/user-guide.md
Original file line number Diff line number Diff line change
Expand Up @@ -1462,7 +1462,7 @@ What it ships (and what it does not):

| Capability | Status |
|------------|--------|
| Presence detection (occupied / empty) | ✅ Trained head — v2 encoder reports 82.3% held-out temporal-triplet acc (v1's "100% on validation" was a single-class recording — retracted, [#882](https://github.com/ruvnet/RuView/issues/882)) |
| Presence detection (occupied / empty) | ⚠️ Trained encoder, degenerate head — the 82.3% is held-out temporal-triplet accuracy (embedding quality), not a presence decision rate. The published head's bias exceeds its weight norm, so it cannot output `absent` for any input ([#1521](https://github.com/ruvnet/RuView/issues/1521)). v1's "100% on validation" was a single-class recording — retracted, [#882](https://github.com/ruvnet/RuView/issues/882) |
| 128-dim CSI embeddings (re-ID, similarity, downstream training) | ✅ Trained encoder |
| Single-person breathing / heart-rate | ⚠️ Server still uses heuristic DSP — model does not replace this yet |
| 17-keypoint full-body pose | 🔬 This HF bundle ships no keypoint head — but real pose weights exist elsewhere; see the tier table below |
Expand Down Expand Up @@ -2143,6 +2143,12 @@ python firmware/esp32-csi-node/provision.py --port COM7 \
--edge-tier 2
```

> **Tier 2 turns off raw CSI.** Tier 0 is raw CSI passthrough with no on-chip DSP; tiers 1 and 2
> compute on the node and ship derived values instead. The room-learning CLI (`enroll`,
> `train-room`, `room-watch`) and anything else that needs per-subcarrier CSI require
> `edge_tier=0` — see `v2/crates/wifi-densepose-cli/src/room.rs`. Provision with `--edge-tier 0`
> when you intend to capture or train, and with `--edge-tier 2` when you want on-node vitals.

Key NVS settings for edge processing:

| NVS Key | Default | What It Controls |
Expand Down