Create app.py

app.py

@@ -0,0 +1,46 @@
+import gradio as gr
+import torch
+from transformers import Wav2Vec2Processor, Wav2Vec2ForSequenceClassification
+import torchaudio
+
+# Load model and processor
+model_name = "Dpngtm/wave2vec2-emotion-recognition"  # Replace with your model's Hugging Face Hub path
+model = Wav2Vec2ForSequenceClassification.from_pretrained(model_name)
+processor = Wav2Vec2Processor.from_pretrained(model_name)
+
+# Define device (use GPU if available)
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model.to(device)
+
+# Preprocessing and inference function
+def recognize_emotion(audio):
+    # Load and resample audio to 16kHz
+    speech_array, sampling_rate = torchaudio.load(audio)
+    if sampling_rate != 16000:
+        resampler = torchaudio.transforms.Resample(sampling_rate, 16000)
+        speech_array = resampler(speech_array)
+    speech_array = speech_array.mean(dim=0).numpy()  # Convert to mono if multi-channel
+
+    # Process input and make predictions
+    inputs = processor(speech_array, sampling_rate=16000, return_tensors="pt", padding=True)
+    inputs = {k: v.to(device) for k, v in inputs.items()}
+    with torch.no_grad():
+        logits = model(**inputs).logits
+    predicted_id = torch.argmax(logits, dim=-1).item()
+
+    # Define emotion labels (use the same order as during training)
+    # Emotion labels mapped to indices
+    emotion_labels = ["angry", "calm", "disgust", "fearful", "happy", "neutral", "sad", "surprised"]
+    return emotion_labels[predicted_id]
+
+# Gradio interface
+interface = gr.Interface(
+    fn=recognize_emotion,
+    inputs=gr.Audio(source="microphone", type="filepath"),
+    outputs="text",
+    title="Emotion Recognition with Wav2Vec2",
+    description="Upload or record audio, and the model will predict the emotion."
+)
+
+# Launch the app
+interface.launch()