Spaces:
Sleeping
Sleeping
File size: 9,411 Bytes
f280f9f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 |
#!/usr/bin/env python3
"""
Madverse Music - Hugging Face Spaces Version
Streamlit app for HF Spaces deployment
"""
import streamlit as st
import torch
import librosa
import tempfile
import os
import time
import numpy as np
# Import the sonics library for model loading
try:
from sonics import HFAudioClassifier
except ImportError:
st.error("Sonics library not found. Please install it first.")
st.stop()
# Global model variable
model = None
# Page configuration
st.set_page_config(
page_title="Madverse Music: AI Music Detector",
page_icon="🎵",
layout="wide",
initial_sidebar_state="expanded"
)
# Custom CSS
st.markdown("""
<style>
.main-header {
background: linear-gradient(90deg, #667eea 0%, #764ba2 100%);
padding: 1rem;
border-radius: 10px;
color: white;
text-align: center;
margin-bottom: 2rem;
}
.result-box {
padding: 1rem;
border-radius: 10px;
margin: 1rem 0;
border-left: 5px solid;
}
.real-music {
background-color: #d4edda;
border-left-color: #28a745;
}
.fake-music {
background-color: #f8d7da;
border-left-color: #dc3545;
}
</style>
""", unsafe_allow_html=True)
@st.cache_resource
def load_model():
"""Load the model with caching for HF Spaces"""
try:
with st.spinner("Loading AI model... This may take a moment..."):
# Use the same loading method as the working API
model = HFAudioClassifier.from_pretrained("awsaf49/sonics-spectttra-alpha-120s")
model.eval()
return model
except Exception as e:
st.error(f"Failed to load model: {str(e)}")
return None
def process_audio(audio_file, model):
"""Process audio file and return classification"""
try:
# Save uploaded file temporarily
with tempfile.NamedTemporaryFile(delete=False, suffix='.wav') as tmp_file:
tmp_file.write(audio_file.read())
tmp_path = tmp_file.name
# Load audio (model uses 16kHz sample rate)
audio, sr = librosa.load(tmp_path, sr=16000)
# Convert to tensor and add batch dimension
audio_tensor = torch.FloatTensor(audio).unsqueeze(0)
# Get prediction using the same pattern as working API
with torch.no_grad():
output = model(audio_tensor)
# Convert logit to probability using sigmoid
probability = torch.sigmoid(output).item()
# Classify: prob < 0.5 = Real, prob >= 0.5 = Fake
if probability < 0.5:
classification = "Real"
confidence = (1 - probability) * 2 # Convert to 0-1 scale
else:
classification = "Fake"
confidence = (probability - 0.5) * 2 # Convert to 0-1 scale
# Calculate duration
duration = len(audio) / sr
# Clean up
os.unlink(tmp_path)
return {
'classification': classification,
'confidence': min(confidence, 1.0), # Cap at 1.0
'probability': probability,
'raw_score': output.item(),
'duration': duration,
'success': True
}
except Exception as e:
# Clean up on error
if 'tmp_path' in locals():
try:
os.unlink(tmp_path)
except:
pass
return {
'success': False,
'error': str(e)
}
def main():
# Header
st.markdown("""
<div class="main-header">
<h1>Madverse Music: AI Music Detector</h1>
<p>Detect AI-generated music vs human-created music using advanced AI technology</p>
</div>
""", unsafe_allow_html=True)
# Sidebar
with st.sidebar:
st.markdown("### About")
st.markdown("""
This AI model can detect whether music is:
- **Real**: Human-created music
- **Fake**: AI-generated music (Suno, Udio, etc.)
**Model**: SpecTTTra-α (120s)
**Accuracy**: 97% F1 score
**Max Duration**: 120 seconds
""")
st.markdown("### Supported Formats")
st.markdown("- WAV (.wav)")
st.markdown("- MP3 (.mp3)")
st.markdown("- FLAC (.flac)")
st.markdown("- M4A (.m4a)")
st.markdown("- OGG (.ogg)")
st.markdown("### Links")
st.markdown("- [Madverse Website](https://madverse.co)")
st.markdown("- [GitHub Repository](#)")
# Load model
model = load_model()
if model is None:
st.error("Model failed to load. Please refresh the page.")
return
st.success("AI model loaded successfully!")
# File upload
st.markdown("### Upload Audio File")
uploaded_file = st.file_uploader(
"Choose an audio file",
type=['wav', 'mp3', 'flac', 'm4a', 'ogg'],
help="Upload an audio file to analyze (max 120 seconds)"
)
if uploaded_file is not None:
# Display file info
st.markdown("### File Information")
col1, col2, col3 = st.columns(3)
with col1:
st.metric("Filename", uploaded_file.name)
with col2:
st.metric("File Size", f"{uploaded_file.size / 1024:.1f} KB")
with col3:
st.metric("Format", uploaded_file.type)
# Audio player
st.markdown("### Preview")
st.audio(uploaded_file)
# Analysis button
if st.button("Analyze Audio", type="primary", use_container_width=True):
try:
with st.spinner("Analyzing audio... This may take a few seconds..."):
# Reset file pointer
uploaded_file.seek(0)
# Process audio
start_time = time.time()
result = process_audio(uploaded_file, model)
processing_time = time.time() - start_time
if not result['success']:
st.error(f"Error processing audio: {result['error']}")
return
# Display results
st.markdown("### Analysis Results")
classification = result['classification']
confidence = result['confidence']
# Result box
if classification == "Real":
st.markdown(f"""
<div class="result-box real-music">
<h3>Result: Human-Created Music</h3>
<p><strong>Classification:</strong> {classification}</p>
<p><strong>Confidence:</strong> {confidence:.1%}</p>
<p><strong>Message:</strong> This appears to be human-created music!</p>
</div>
""", unsafe_allow_html=True)
else:
st.markdown(f"""
<div class="result-box fake-music">
<h3>Result: AI-Generated Music</h3>
<p><strong>Classification:</strong> {classification}</p>
<p><strong>Confidence:</strong> {confidence:.1%}</p>
<p><strong>Message:</strong> This appears to be AI-generated music!</p>
</div>
""", unsafe_allow_html=True)
# Detailed metrics
with st.expander("Detailed Metrics"):
col1, col2, col3 = st.columns(3)
with col1:
st.metric("Confidence", f"{confidence:.1%}")
with col2:
st.metric("Probability", f"{result['probability']:.3f}")
with col3:
st.metric("Processing Time", f"{processing_time:.2f}s")
if result['duration'] > 0:
st.metric("Duration", f"{result['duration']:.1f}s")
st.markdown("**Interpretation:**")
st.markdown("""
- **Probability < 0.5**: Classified as Real (human-created)
- **Probability ≥ 0.5**: Classified as Fake (AI-generated)
- **Confidence**: How certain the model is about its prediction
""")
except Exception as e:
st.error(f"Error processing audio: {str(e)}")
# Footer
st.markdown("---")
st.markdown("""
<div style="text-align: center; color: #666;">
<p>Powered by <strong>Madverse Music</strong> | Built with Streamlit & PyTorch</p>
<p>This tool is for research and educational purposes. Results may vary depending on audio quality.</p>
</div>
""", unsafe_allow_html=True)
if __name__ == "__main__":
main() |