56 lines
1.7 KiB
Markdown
56 lines
1.7 KiB
Markdown
# demo-ai-face-detector
|
|
|
|
CUTOS 4.0 Vue 3/Vite LWA demonstrating local browser-side face detection with `@cutos/ai-face-detector`.
|
|
|
|
No Device Provider or Gateway service is required. The LWA needs camera permission and packages the MediaPipe model/WASM resources for offline use.
|
|
|
|
## Behaviour
|
|
|
|
- Starts the user-facing camera automatically.
|
|
- Uses the source camera ratio in landscape and a centre-cropped 1:1 preview in portrait.
|
|
- Guides the user with a circular capture area and a moving confidence box.
|
|
- Requires one face, confidence ≥ 0.90, face size ≥ 12%, and horizontal centering.
|
|
- Requires the accepted face position and size to remain stable for about two seconds.
|
|
- Locks the stable face image and stops inference while keeping the camera preview visible.
|
|
- Restarts detection automatically after ten seconds, or immediately when `Detect again` is selected.
|
|
- Displays the locked face in the lower-right corner and keeps Detection Result and Activity in two equal cards.
|
|
|
|
## Development with the sibling SDK
|
|
|
|
```sh
|
|
cd ../sdk
|
|
npm install
|
|
npm run build
|
|
|
|
cd ../demo-ai-face-detector
|
|
npm install
|
|
npm run dev
|
|
```
|
|
|
|
Both `dev` and the CUTOS build copy version-matched model assets through the SDK's asset command.
|
|
|
|
## Build and package
|
|
|
|
```sh
|
|
npx cutos lwa build
|
|
```
|
|
|
|
The generated package is:
|
|
|
|
```text
|
|
release/demo-ai-face-detector-v<version>.lwa
|
|
```
|
|
|
|
The LWA version comes from `public/config.json`; `package.json` intentionally remains `0.0.0`.
|
|
|
|
## Upload and publish
|
|
|
|
```sh
|
|
npx cutos login --username office
|
|
npx cutos lwa upload
|
|
npx cutos lwa list demo-ai-face-detector
|
|
npx cutos lwa publish --lwa <content-id> --device <device-id>
|
|
```
|
|
|
|
For a release build, depend on the published npm SDK rather than `file:../sdk`.
|