Delete app.py

app.py

@@ -1,53 +0,0 @@
-import gradio as gr, numpy as np, librosa, tensorflow as tf, os, json
-
-SR, N_SAMPLES, N_MELS, HOP, NFFT = 16000, 16000, 64, 160, 400
-PAD_T = 100
-LABELS = ['yes','no','up','down','left','right','on','off','stop','go','unknown','silence']
-
-model = tf.keras.models.load_model("speech_command_cnn")  # put model files in Space or load from Hub
-
-def load_and_feat(path):
-    y, _ = librosa.load(path, sr=SR)
-    if len(y) > N_SAMPLES: y = y[:N_SAMPLES]
-    if len(y) < N_SAMPLES: y = np.pad(y, (0, N_SAMPLES-len(y)))
-    S = librosa.feature.melspectrogram(y=y, sr=SR, n_fft=NFFT, hop_length=HOP, n_mels=N_MELS, fmax=8000)
-    logS = librosa.power_to_db(S, ref=np.max)
-    logS = (logS - logS.mean()) / (logS.std()+1e-9)
-    f = logS[..., np.newaxis]
-    T = f.shape[1]
-    if T < PAD_T: f = np.pad(f, ((0,0),(0,PAD_T-T),(0,0)))
-    else: f = f[:, :PAD_T, :]
-    return f.astype(np.float32)
-
-INTENTS = {
-    'yes': 'confirm',
-    'no': 'cancel',
-    'stop': 'pause',
-    'go': 'start',
-    'on': 'turn_on',
-    'off': 'turn_off',
-    'up': 'volume_up',
-    'down': 'volume_down',
-    'left': 'move_left',
-    'right': 'move_right'
-}
-
-def predict(audio_path):
-    f = load_and_feat(audio_path)
-    p = model.predict(f[np.newaxis,...], verbose=0)[0]
-    top_idx = np.argsort(-p)[:3]
-    top = [(LABELS[i], float(p[i])) for i in top_idx]
-    label = LABELS[int(np.argmax(p))]
-    intent = INTENTS.get(label, 'none')
-    return {k: float(v) for k,v in top}, intent
-
-demo = gr.Interface(
-    fn=predict,
-    inputs=gr.Audio(sources=["microphone","upload"], type="filepath"),
-    outputs=[gr.Label(num_top_classes=3), gr.Textbox(label="Intent")],
-    title="Voice Command → Intent",
-    description="Speak a short command: yes/no/stop/go/up/down/left/right/on/off"
-)
-
-if __name__ == "__main__":
-    demo.launch()