What does it do?

Sound AI is a audio training platform whose purpose is to provide the users with a powerful model that solves the analyzing and classifying audio data. This version supports classification of sound between 2 classes i.e binary classification problems. Once the training is done, the model file generated will be shared with the user for downloading and to use in his/her application for free.

The user only needs to provide the platform with the unprocessed audio data. This simple, yet sufficient platform has enough to augment your data and generate your model. A bunch of audio files and an email ID is all you need to get started.

Any binary audio dataset, however imbalanced they might be, Sound AI is a one stop solution to train them. Built on top of Google Forms and Google Drive, this subtle yet supple platform hides a powerful classifier that might surprise you. Once the user provides the audio data, the platform takes care of length unification, data augmentation and model generation. You can use the model in your script like this.

Alternatively, SoundAI also extends it's API to a prebuilt model that you can use in your applications.

Looking to build something on your own? Check this blog out.

Key Challenges

Since the app is free to use, one of the challenges was to reduce the development cost. The google forms as and google drive as backend enabled this. They are simple to use and cost free and more importantly, were built to the need within half a day.

Another challenge was to make an app that does not require much from the end user beyond the data collection and minimalistic noise removal.

Technologies used

Python, Tensorflow, AWS, Google forms