About

DeepAudioX is a PyTorch-based library that provides simple, flexible pipelines for audio classification using pretrained audio foundation models as feature extractors.

It is designed to let users train, evaluate, and run inference on custom audio datasets with only a few lines of code, while still allowing advanced customization when needed.

Key Features

• 🔊 Pretrained audio backbones for feature extraction

• 🧠 Modular pooling strategies (e.g. GAP, SimPool, EfficientProbing)

• 🧩 Custom classifier heads for downstream audio classification

• 🚀 High-level training, evaluation, and inference APIs

• 🔁 Fully PyTorch-native and extensible

• 📦 Clean integration with existing PyTorch workflows

Quick Start

With DeepAudio-X, you can build and train an audio classifier in just a few lines of code, leveraging pretrained backbones for state-of-the-art performance. Suppose you have a dataset organized in a directory structure where each subdirectory corresponds to a class label and contains audio files for that class as follows:

path/to/data/
    class_1/
        audio_1.wav
        audio_2.wav
        ...
    class_2/
        audio_3.wav
        audio_4.wav
        ...
    ...

Then you can use the following minimal code to create, train, and evaluate an audio classifier:

from deepaudiox import AudioClassifier, Evaluator, Trainer
from deepaudiox import audio_classification_dataset_from_dir
from deepaudiox import get_class_mapping_from_dir

# 1) Build a dataset from a folder structure of class subdirectories
class_mapping = get_class_mapping_from_dir(root_dir="path/to/data")
dataset = audio_classification_dataset_from_dir(
    root_dir="path/to/data",
    sample_rate=16_000,
    class_mapping=class_mapping,
)

# 2) Create a classifier with a pretrained backbone
classifier = AudioClassifier(
    num_classes=len(class_mapping),
    backbone="beats",
    sample_rate=16_000,
    pretrained=True,
    freeze_backbone=True,
)

# 3) Train
trainer = Trainer(
    train_dset=dataset,
    model=classifier,
    validation_dset=dataset,  # replace with a real validation set
    batch_size=16,
    epochs=5,
)

trainer.train()

classifier = AudioClassifier.from_checkpoint("checkpoint.pt")  # Load model with config restored

# 4) Evaluate on a test set
evaluator = Evaluator(
    test_dset=dataset,  # replace with a real test set
    model=classifier,
    batch_size=16,
)

evaluator.evaluate()