Music 4 You

Inspiration

We like how Spotify has tailored playlists for a multitude of moods, and their recommendation system for music genres. However, we think it would be so much better for Spotify to automatically provide music for users' mood, rather than them having to pick it out themselves. The whole purpose of this is to make users feel better!

What it does

It uses a pre-trained model to detect the facial expression, age and gender of users. Using information about facial expression, the web application recommends playlists for the user's mood.

How we built it

We built it using a pre-trained model with Convolutional Neural Network [UTKFace Dataset from Kaggle] to predict the facial expression, age and gender of the user. We then used the Spotify API to recommend Spotify playlists embedded right into our website.

Challenges we ran into

We struggled in the web development aspect in implementing flask, and this took up a bulk of our time. We would've used the time to improve the UI further and to apply more Deep Learning networks.

Accomplishments that we're proud of

We spent some time brainstorming this idea and hence, we are proud of it! We are also glad to have picked up new coding skills through this challenge.

What we learned

We learnt how to make machine learning models interactive with users in a web environment, and how to implement API better. We also faced some challenges implementing the user interface and we know that we will be able to do it better in the future.

What's next for Music 4 You

Music 4 You could be further improved by

Using the age and gender of users to predict playlists for these users better.
Include individual song recommendations that could better suit the users as they can be individually catered to their preferences based on the music genres and the age groups of users.
We would also like to allow the option of using webcam to for facial detection, for which libraries are readily available already!
With more time, we can also improve on the user experience with the interface
We would also like to include YouTube video predictions as well to maximise the usage of this system.