Add NN architecture + live browser demo (wasm-demo-tester) to card

README.md

@@ -21,9 +21,10 @@ datasets:
 
 Hey-Edge is an audio keyword-spotting wake-word model trained with Edge Impulse to detect the phrase "hey edge" from 16 kHz microphone audio.
 
-The model was trained using synthetic and augmented audio and exported as an Edge Impulse C++ library for embedded and TinyML deployment.
+The model was trained using synthetic and augmented audio and exported as an Edge Impulse C++ library for embedded and TinyML deployment. A WebAssembly (browser) export of the same model also runs live in the [wasm-demo-tester Space](https://huggingface.co/spaces/edgeimpulse/wasm-demo-tester).
 
-Live Edge Impulse project: https://studio.edgeimpulse.com/public/1052106/live
+- Live Edge Impulse project: https://studio.edgeimpulse.com/public/1052106/live
+- Try it in the browser (mic): https://huggingface.co/spaces/edgeimpulse/wasm-demo-tester
 
 ## Model details
 
@@ -62,6 +63,16 @@ This model is not intended for speaker identification, speech recognition, trans
 | Data type | Synthetic and augmented audio |
 | Audio sample rate | 16 kHz |
 
+## Neural network architecture
+
+Transfer-learning keyword-spotting head (`keras-transfer-kws`) on MFE audio features.
+
+| Layer | Detail |
+|---|---|
+| Input layer | 3,960 features |
+| Backbone | MobileNetV2 0.35 (no final dense layer, 0.1 dropout) |
+| Output layer | 3 classes (`background_noise`, `hey_edge`, `unknown`) |
+
 ## Validation performance
 
 | Metric | Value |