A modified implementation of Fader Networks optimized for manipulating timbral adjectives in audio files using Mel Scale Spectrograms.
- Audio Domain: Adapted the original architecture to process audio features instead of image data.
- Target Qualities: Trained using the NSynth Dataset to control specific "Note Qualities": Bright, Dark, Distortion, and Percussive.
- Modern Compatibility: Updated to support PyTorch v2.0.1+.
- Execution: Optimized for CPU inference (CUDA support is currently disabled/not ported).
This is a personal fork for experimentation. All rights and original architecture credits belong to the original authors; PyTorch implementation of Fader Networks (NIPS 2017).
Fader Networks: Manipulating Images by Sliding Attributes - G. Lample, N. Zeghidour, N. Usunier, A. Bordes, L. Denoyer, M'A. Ranzato
@inproceedings{lample2017fader,
title={Fader Networks: Manipulating Images by Sliding Attributes},
author={Lample, Guillaume and Zeghidour, Neil and Usunier, Nicolas and Bordes, Antoine and DENOYER, Ludovic and others},
booktitle={Advances in Neural Information Processing Systems},
pages={5963--5972},
year={2017}
}