DeepFense is a modular, configuration-driven framework for deepfake audio detection. Mix and match frontends, backends, and loss functions via YAML -- no code changes needed.
| Step | Guide | Description |
|---|---|---|
| 1 | Installation | Set up your environment |
| 2 | Quick Start | Train your first model in 5 minutes |
| 3 | Full Tutorial | Every config parameter explained |
| Guide | Description |
|---|---|
| Architecture | How DeepFense works internally |
| Configuration | All YAML parameters |
| Library Usage | Use DeepFense as a Python library |
| HuggingFace Hub | Download datasets & pretrained models |
| CLI Reference | Command-line interface |
| Pipeline Flow | Complete data-to-evaluation pipeline |
| Data Transforms | All padding, cropping, and augmentation options |
| Component | Description |
|---|---|
| Frontends | Wav2Vec2, WavLM, HuBERT, MERT, EAT |
| Backends | AASIST, ECAPA-TDNN, Nes2Net, RawNet2, MLP, TCM |
| Losses | CrossEntropy, OC-Softmax, AM-Softmax, A-Softmax |
| Augmentations | RawBoost, RIR, Codec, Noise, SpeedPerturb, ... |
| Optimizers & Schedulers | Adam, SGD, CosineAnnealing, StepLR, ... |
| Guide | Description |
|---|---|
| Extending DeepFense | Quick reference for all component types |
| Adding Frontends | Create custom feature extractors |
| Adding Backends | Create custom classifiers |
| Adding Losses | Create custom loss functions |
| Adding Datasets | Create custom datasets |
| Adding Augmentations | Create custom augmentations |
| Adding Optimizers | Add custom optimizers |
| Adding Schedulers | Add custom schedulers |
| Adding Metrics | Add custom evaluation metrics |
| Training Workflow | Detailed training loop explanation |
| Training with CLI | CLI-based training |
| Inference | Testing and deployment |
New to DeepFense? Start here: Installation -> Quick Start -> Full Tutorial