Gismart at Datathon’18 presenting ML challenge

Gismart has partnered with Datathon, Big Data & Data Science hackathon. Datathon is a 6-day competition where talented teams come up with a working solution for real-life business challenges of the top companies operating in finance, telecom, consumption, mobile and gaming technology domains with the help of data science.

As a music tech hackathon partner, we have set a task to improve the clarity of vocal recordings in Karaoke by Gismart app using machine learning (ML) and artificial intelligence (AI) technology. Karaoke by Gismart, the first Karaoke app with AR masks lets users create video and audio recording. Due to imperfect acoustic conditions vocal recordings are often interrupted with external noises. The hackathon participants were set to provide a solution that diminishes or eliminates external noise with ML and AI technology.

Datathon participants worked with provided audio recordings that had multiple disruptive noises – constant background noise, on and off noises, the instrumental noise from a song itself when the vocalist is singing without headphones, etc.

“The quality of audio recording has been one of the main issues in Karaoke app. We’ve recently established in-house Research and Development Department in the field of voice and sounds processing using ML and AI technologies where we look for the best tech solutions that ultimately improve the product and end-user experience. Datathon was an opportunity for us to meet the community, share our knowledge and see solutions provided by the teams,” – Sergey Shchegrikovich, Director of Engineering at Gismart.

Gismart has chosen Enigma team as the winner of Gismart challenge.