| | --- |
| | license: mit |
| | --- |
| | A deep learning model for audio classification, designed to detect AI-generated music. |
| |
|
| | Model architecture : EfficientNetB0 (Transfer Learning) |
| |
|
| | Framework: TensorFlow / Keras |
| |
|
| | Task: Binary Classification (AI Music vs. Real Music) |
| |
|
| | Input: Mel Spectrograms (224x224) generated from audio files. |
| |
|
| | Training Data: Trained on a balanced dataset of thousands of |
| | modern songs from sources like the Free Music Archive, NCS, |
| | and the Suno AI dataset on Kaggle, all standardized to a |
| | consistent bitrate. |
