Phone recognition

Example 31 uses an optional object detector. Point at another phone and enable «Mostrar detección y confianza» to see the detected box and confidence. Kai follows the box on screen. Inference runs locally in a worker.

Install the phone detector (example 31)

The detector is optional and distributed separately from webar-facil.min.js. To add it to your project, copy 31-reconocer-celular.html, the entire phone-detector/ folder alongside the HTML, and the GLB model you want to show.

• Keep demo.js, worker.js, tf.min.js, coco-ssd.min.js, model.json, and all five files from group1-shard1of5 through group1-shard5of5 together. The weights manifest uses those exact names. Include LICENSE.txt and keep README.md for reference.
• In the HTML, point the first <script> to your copy of webar-facil.min.js and keep <script type="module" src="./phone-detector/demo.js"></script>. If you change the GLB or folder layout, update the path in demo.js. The model path resolves relative to demo.js, not the HTML.
• Serve the files over HTTPS or use localhost during development. The worker loads the weights from its own folder and processes camera frames locally.

Apps that only use markers or SLAM can use the SDK without copying the detector.