Creative Programming and Computing
A. Y. 2022-2023
- Giorgio Bonizzi (https://github.com/Embodynglmp)
- Viola Negroni (https://github.com/violangg)
- Enrico Roncuzzi (https://github.com/enricoroncuzzi)
- Enrico Francesco Valente (https://github.com/enricovalente23)
github.com/violangg/CPAC22-Project
Here we present a method for creating a portrait of a person based on their listening habits. The main goal is to create a unique and personalized depiction of the person that reflects their musical preferences. Our method is a novel approach to creating portraits and has potential applications in the fields of art, music, and personalization.
Do you really know yourself? Everyone knows your face, but how much does it really tell about your inner-self? If there is an activity in which you can be totally sincere with yourself is listening to music: a private space in which you can confront your feelings.
The goal of the project is therefore to give the user the possibility to look at him/herself from a new perspective, in which the traits of his/her face change depending on his/her musical taste.
The core of the project, which we aim to implement within the duration of the hackaton as a prototype, is to analyze the user's listening habits exploiting Spotify APIs and find his/her place in the "spectrum" of an Emotional Model, which is build starting from the features extracted from music. In order to build a personalized virtual palette, each emotion on the spectrum will be associated to specific visual features, such as specific colors (or even shapes and painting styles).
A neural network (perhaps a CNN such as the VGG16) will then be exploited: taking as inputs the results coming from this last process and an image of the user's face, a portrait will be "painted" using a style which is strictly linked to the user's musical personality.
The ultimate goal is to increase the complexity of the emotional model, adding layers of variability and exploiting new features extracted from the music library. As a direct consequence, the next improvement will be the implementation of more artistic features, allowing to generate a richer and more personalized output.
It could be interesting as well to realize a 3D mesh of the user's face instead of just realizing a bidimensional painting, or even to include the possibility to modify, along with the style of the portrait, also the facial expression of the user (e.g make him/her smile if the main detected emotion is happiness). Also, a desirable side upgrade could be to make the interaction between the user and the application easier, without having to request every time a token to use Spotify APIs.
