Create appp.py

appp.py

@@ -0,0 +1,81 @@
+import tensorflow as tf
+import joblib
+import numpy as np
+import gradio as gr
+from scipy.io import wavfile
+import os
+
+# Load model and label encoder
+model = tf.keras.models.load_model("animal_sound_cnn.keras")
+label_encoder = joblib.load("label_encoder.joblib")
+
+def preprocess_audio(audio_path, target_shape=(64, 64)):
+    """
+    Simplified audio preprocessing using only numpy/scipy
+    """
+    try:
+        # 1. Read WAV file
+        sr, y = wavfile.read(audio_path)
+        
+        # Convert to mono if stereo
+        if len(y.shape) > 1:
+            y = y.mean(axis=1)
+            
+        # 2. Simple spectrogram using STFT
+        f, t, spec = tf.signal.stft(
+            y,
+            frame_length=256,
+            frame_step=128,
+            fft_length=256
+        )
+        spectrogram = np.abs(spec)
+        
+        # 3. Resize to target dimensions
+        spectrogram = tf.image.resize(
+            tf.expand_dims(spectrogram, -1),
+            target_shape
+        ).numpy()
+        
+        # 4. Normalize and add batch dimension
+        spectrogram = (spectrogram - spectrogram.min()) / (spectrogram.max() - spectrogram.min())
+        return spectrogram[np.newaxis, ..., np.newaxis].astype(np.float32)
+        
+    except Exception as e:
+        print(f"Preprocessing error: {str(e)}")
+        return None
+
+def predict(audio_path):
+    try:
+        # 1. Preprocess audio
+        spectrogram = preprocess_audio(audio_path)
+        if spectrogram is None:
+            return "Error: Could not process audio"
+            
+        # 2. Debug log input shape
+        print(f"Input shape: {spectrogram.shape}")
+        
+        # 3. Predict
+        pred = model.predict(spectrogram)
+        animal = label_encoder.inverse_transform([np.argmax(pred)])[0]
+        
+        return animal
+        
+    except Exception as e:
+        return f"Prediction error: {str(e)}"
+
+# Minimal requirements.txt needed:
+# tensorflow>=2.16.0
+# scikit-learn
+# joblib
+# numpy
+# gradio
+# scipy
+
+gr.Interface(
+    fn=predict,
+    inputs=gr.Audio(type="filepath"),
+    outputs="label",
+    title="Animal Sound Classifier",
+    description="Upload a short audio clip (3-5 seconds) of an animal sound",
+    examples=["example.wav"] if os.path.exists("example.wav") else None
+).launch()