Dylan Harness
Sqwak
Build and deploy audio classifiers with a few clicks
Sqwak uses the fast fourier transform and other features from the uploaded audio files to train a neural network to classify audio files. The neural network is trained using Tensorflow. The model is pickled and stored in a database. When an inference request is made, the model is loaded from the database and used to predict the class of the audio file.
Sqwak lets you create an audio classifier without writing any code! To train your classifier, create some bins (such as 'clanking' 'quacking' and 'rock music') and upload your music into the appropriate bin. After that, click the train button and voila, your classifier is now available at an api endpoint. You can now POST to that endpoint with completely new sounds your classifier has never seen before, and it'll attempt to label it.
How it Works
- Create a class you want to detect (example: a dog barking)
- Upload audio files of that sound
- Click "Train" and Sqwak handles all the audio parsing and training for you.
- Publish your model to an API endpoint with 1 click.
- Access your personal endpoint from your product to classify new audio files!